diff --git a/CLAUDE.md b/CLAUDE.md
index 7536204..ee16581 100644
--- a/CLAUDE.md
+++ b/CLAUDE.md
@@ -1526,6 +1526,68 @@ of it**, measured on trees verified identical beforehand. What made that safe to
 it was that the README transcripts and the figure gate would have said
 otherwise if it were not.
 
+### `strict=` is the whole point, and ruff's own fix takes it away
+
+`zip()` without `strict=` truncates to the shorter argument and says nothing.
+The ratchet recorded 41, and the obvious move is `ruff --fix --unsafe-fixes`.
+**It writes `strict=False`** -- the current behaviour, spelled out loud. That
+takes the count to zero, looks exactly like the job being done, and preserves
+every defect the rule exists to expose. Check *which* keyword a mechanical fix
+inserted before believing the number it produced.
+
+Done by hand instead, and the count was the least interesting part:
+
+- **58 sites, not 41.** The other 17 are in `notebooks/`, which is not in
+  `SOURCE_DIRS` in `tests/test_lint_debt_only_shrinks.py`, so no ruff guard here
+  had ever read them -- in the three notebooks a student is most likely to open.
+  Same shape as `KNOWN_NON_EXAMPLE_CHAPTER_FILES`: **a file outside the sweep's
+  scope is a file nothing sweeps**, and that scope is a tuple somebody wrote
+  once.
+- **One real defect**, in `ch5_fingerprinting/example_comparison.py`, where
+  `colors_box` held five shades against six methods. The sixth box kept
+  matplotlib's default blue among five pastels and read as deliberately
+  highlighted. Nothing raised, nothing left a range, no printed number moved --
+  the sixth box simply never got its colour, which only opening the PNG shows.
+- **One site where `strict=True` is wrong.** `zip(scales, scales[1:])` is the
+  consecutive-pairs idiom and *depends* on the truncation, so strict raises on
+  every call; it is `itertools.pairwise` now. `zip(xs, itertools.count())` and
+  `zip(xs, cycle(ys))` are the same family. **The tell is a second argument
+  derived from the first**, and a grep for `[1:]` finds only the spelling you
+  thought of -- read the hunks.
+
+**The entry is deleted from BASELINE rather than set to 0, and the deletion is
+the guard**: `test_no_rule_has_more_findings_than_it_did` fails on any rule not
+listed, so one new unguarded `zip()` turns it red. Verified by adding one and
+watching `B905: 1 findings, new`, not by reading the assertion.
+
+**A green sweep only means something where the line runs, so that was measured
+too. 41 of the 42 `.py` sites execute**, and the one that does not is the
+interesting entry. 34 run under the examples, the generators and two flag-gated
+paths a plain run never reaches -- ch4's `--data` branch and ch7's `--animate`
+callback, which between them hold two sites nothing else touches. Six more sit
+inside test bodies and one in `core/fingerprinting`, all seven confirmed rather
+than assumed. The last, ch8's unobservable-mode print, is behind
+`if n_unobservable > 0`, which the fixes-aided configuration never enters. It
+is equal-length by construction -- `Vt` from `svd(..., full_matrices=True)` is
+square, so a mode column is `n_states` long and `state_names` comes from the
+same analysis -- and its identical twin in the odometry-only branch executes.
+The 17 notebook sites are covered by `tests/docs/test_notebooks_run.py`.
+
+**The measurement lied twice before it said anything true**, both times looking
+exactly like "this code never runs":
+
+- `[tool.coverage.run] source = ["ipin_examples", "core"]` in `pyproject.toml`
+  means a plain `coverage run` over a chapter example measures `core/` and
+  **not the example**, so the first cross-reference reported 42 of 43 sites as
+  never loaded. Pass `--source` explicitly when measuring something the project
+  does not normally measure.
+- `coverage combine` **starts clean unless you pass `--append`**. Running it a
+  second time discarded a finished 16-step sweep, and the same script then
+  reported 1 site executed where it had reported 32.
+
+Both were caught the same way: a number moved while the code did not. That is
+the fifth harness in this file to report the thing it could not read as broken.
+
 ## Parallel sessions
 
 Several agents often work this repo at once, on separate worktrees off `main`.
diff --git a/ch3_estimators/example_comparison.py b/ch3_estimators/example_comparison.py
index 4bb3b1a..5fcfd33 100644
--- a/ch3_estimators/example_comparison.py
+++ b/ch3_estimators/example_comparison.py
@@ -656,7 +656,7 @@ def main():
     ax.grid(True, alpha=0.3, axis="y")
 
     # Add value labels on bars
-    for bar, n_evals in zip(bars, evaluations):
+    for bar, n_evals in zip(bars, evaluations, strict=True):
         height = bar.get_height()
         ax.text(
             bar.get_x() + bar.get_width() / 2.0,
diff --git a/ch4_rf_point_positioning/example_aoa_positioning.py b/ch4_rf_point_positioning/example_aoa_positioning.py
index a207a80..af80e5f 100644
--- a/ch4_rf_point_positioning/example_aoa_positioning.py
+++ b/ch4_rf_point_positioning/example_aoa_positioning.py
@@ -386,7 +386,7 @@ def visualize_aoa_geometry():
     #
     # Negating gives the anchor-to-agent direction, so the rays now meet at
     # the target the way an AOA geometry figure is meant to show.
-    for i, (anchor, psi) in enumerate(zip(anchors, aoa_noisy)):
+    for i, (anchor, psi) in enumerate(zip(anchors, aoa_noisy, strict=True)):
         # In ENU: psi is from North (+y), so the agent lies from the anchor at
         # E-component = -sin(psi), N-component = -cos(psi)
         line_length = 15
diff --git a/ch4_rf_point_positioning/example_comparison.py b/ch4_rf_point_positioning/example_comparison.py
index 49c6a6a..82e9afa 100644
--- a/ch4_rf_point_positioning/example_comparison.py
+++ b/ch4_rf_point_positioning/example_comparison.py
@@ -778,7 +778,7 @@ def plot_dataset_results(results: Dict, output_file: str = None):
         results["gdop"]["AOA"],
     ]
     bp = ax3.boxplot(gdop_data, tick_labels=["TOA", "TDOA", "AOA"], patch_artist=True)
-    for patch, color in zip(bp["boxes"], ["blue", "red", "green"]):
+    for patch, color in zip(bp["boxes"], ["blue", "red", "green"], strict=True):
         patch.set_facecolor(color)
         patch.set_alpha(0.6)
     ax3.set_ylabel("GDOP")
@@ -934,7 +934,7 @@ def plot_inline_comparison(noise_levels, results):
 
     # 1. RMSE vs Noise
     ax1 = axes[0, 0]
-    for method, color in zip(methods, colors):
+    for method, color in zip(methods, colors, strict=True):
         rmse_values = [
             np.sqrt(np.mean(e**2)) if len(e) > 0 else np.nan for e in results[method]
         ]
@@ -950,7 +950,7 @@ def plot_inline_comparison(noise_levels, results):
     # 2. Error CDF
     ax2 = axes[0, 1]
     noise_idx = 2
-    for method, color in zip(methods, colors):
+    for method, color in zip(methods, colors, strict=True):
         errors = results[method][noise_idx]
         if len(errors) > 0:
             sorted_errors = np.sort(errors)
@@ -968,7 +968,7 @@ def plot_inline_comparison(noise_levels, results):
     data = [results[m][noise_idx] for m in methods if len(results[m][noise_idx]) > 0]
     labels = [m for m in methods if len(results[m][noise_idx]) > 0]
     bp = ax3.boxplot(data, tick_labels=labels, patch_artist=True, showfliers=False)
-    for patch, color in zip(bp["boxes"], colors[: len(data)]):
+    for patch, color in zip(bp["boxes"], colors[: len(data)], strict=True):
         patch.set_facecolor(color)
         patch.set_alpha(0.6)
     ax3.set_ylabel("Position Error (m)")
@@ -986,7 +986,7 @@ def plot_inline_comparison(noise_levels, results):
     ax4 = axes[1, 1]
     total_points = 50
     dashes = ["-", "--", "-.", ":"]
-    for method, color, dash in zip(methods, colors, dashes):
+    for method, color, dash in zip(methods, colors, dashes, strict=True):
         rates = [len(e) / total_points * 100 for e in results[method]]
         ax4.plot(
             noise_levels,
diff --git a/ch5_fingerprinting/example_classification.py b/ch5_fingerprinting/example_classification.py
index 65c30bc..1f34efc 100644
--- a/ch5_fingerprinting/example_classification.py
+++ b/ch5_fingerprinting/example_classification.py
@@ -442,7 +442,7 @@ def evaluate_hierarchical_localization(db: FingerprintDatabase, rng=None):
     ax.set_title("RMSE Comparison", fontsize=12)
     ax.grid(True, alpha=0.3, axis="y")
     # Add values on bars
-    for bar, rmse in zip(bars, rmses):
+    for bar, rmse in zip(bars, rmses, strict=True):
         height = bar.get_height()
         ax.text(
             bar.get_x() + bar.get_width() / 2,
diff --git a/ch5_fingerprinting/example_comparison.py b/ch5_fingerprinting/example_comparison.py
index b8093a6..19b7ec6 100644
--- a/ch5_fingerprinting/example_comparison.py
+++ b/ch5_fingerprinting/example_comparison.py
@@ -289,7 +289,7 @@ def evaluate_scenario(scenario_name, db, queries, true_locs, floor_id=None):
     method_name = "NN (Euclidean)"
     print(f"  {method_name}...", end=" ", flush=True)
     errors, times = [], []
-    for query, true_loc in zip(queries, true_locs):
+    for query, true_loc in zip(queries, true_locs, strict=True):
         t_start = time.perf_counter()
         est_loc = nn_localize(query, db, metric="euclidean", floor_id=floor_id)
         t_end = time.perf_counter()
@@ -315,7 +315,7 @@ def evaluate_scenario(scenario_name, db, queries, true_locs, floor_id=None):
     method_name = "k-NN (k=3)"
     print(f"  {method_name}...", end=" ", flush=True)
     errors, times = [], []
-    for query, true_loc in zip(queries, true_locs):
+    for query, true_loc in zip(queries, true_locs, strict=True):
         t_start = time.perf_counter()
         est_loc = knn_localize(
             query,
@@ -356,7 +356,7 @@ def evaluate_scenario(scenario_name, db, queries, true_locs, floor_id=None):
     method_name = "MAP"
     print(f"  {method_name}...", end=" ", flush=True)
     errors, times = [], []
-    for query, true_loc in zip(queries, true_locs):
+    for query, true_loc in zip(queries, true_locs, strict=True):
         t_start = time.perf_counter()
         est_loc = map_localize(query, model_bayes, floor_id=floor_id)
         t_end = time.perf_counter()
@@ -382,7 +382,7 @@ def evaluate_scenario(scenario_name, db, queries, true_locs, floor_id=None):
     method_name = "Posterior Mean"
     print(f"  {method_name}...", end=" ", flush=True)
     errors, times = [], []
-    for query, true_loc in zip(queries, true_locs):
+    for query, true_loc in zip(queries, true_locs, strict=True):
         t_start = time.perf_counter()
         est_loc = posterior_mean_localize(query, model_bayes, floor_id=floor_id)
         t_end = time.perf_counter()
@@ -408,7 +408,7 @@ def evaluate_scenario(scenario_name, db, queries, true_locs, floor_id=None):
     method_name = "Post.Mean (k=10)"
     print(f"  {method_name}...", end=" ", flush=True)
     errors, times = [], []
-    for query, true_loc in zip(queries, true_locs):
+    for query, true_loc in zip(queries, true_locs, strict=True):
         t_start = time.perf_counter()
         est_loc = posterior_mean_localize(
             query, model_bayes, floor_id=floor_id, top_k=10
@@ -443,7 +443,7 @@ def evaluate_scenario(scenario_name, db, queries, true_locs, floor_id=None):
     method_name = "Linear Regression"
     print(f"  {method_name}...", end=" ", flush=True)
     errors, times = [], []
-    for query, true_loc in zip(queries, true_locs):
+    for query, true_loc in zip(queries, true_locs, strict=True):
         t_start = time.perf_counter()
         est_loc = model_lr.predict(query)
         t_end = time.perf_counter()
@@ -604,8 +604,19 @@ def main():
     ax4 = plt.subplot(3, 3, 4)
     error_data = [r["errors"] for r in all_results["Baseline"]]
     bp = ax4.boxplot(error_data, tick_labels=methods, patch_artist=True)
-    colors_box = ["lightblue", "lightcyan", "lightcoral", "lightsalmon", "lightgreen"]
-    for patch, color in zip(bp["boxes"], colors_box):
+    # One light shade per entry of `colors` above, in the same order. It was a
+    # shade short, so the sixth method's box kept matplotlib's default facecolor
+    # and read as deliberately highlighted among five pastel ones. strict=True
+    # below is what stops the two lists drifting apart again.
+    colors_box = [
+        "lightblue",
+        "lightcyan",
+        "lightcoral",
+        "lightsalmon",
+        "lightgreen",
+        "plum",
+    ]
+    for patch, color in zip(bp["boxes"], colors_box, strict=True):
         patch.set_facecolor(color)
     ax4.set_ylabel("Positioning Error (m)")
     ax4.set_title("Error Distribution (Baseline)")
@@ -641,7 +652,7 @@ def main():
     # actual finding, so it should not be what breaks the labelling. Colours
     # match the per-query cost panel above, so a method reads the same in both.
     ax6 = plt.subplot(3, 3, 6)
-    for r, color in zip(all_results["Baseline"], colors):
+    for r, color in zip(all_results["Baseline"], colors, strict=True):
         ax6.scatter(
             r["ops_per_query"],
             r["rmse"],
diff --git a/ch5_fingerprinting/example_deterministic.py b/ch5_fingerprinting/example_deterministic.py
index 867bf5f..797f030 100644
--- a/ch5_fingerprinting/example_deterministic.py
+++ b/ch5_fingerprinting/example_deterministic.py
@@ -77,7 +77,7 @@ def generate_test_queries(db, n_queries=100, floor_id=None, noise_std=0.0, seed=
     # Generate fingerprints by interpolating from nearby RPs
     query_fingerprints = []
 
-    for i, (true_loc, fid) in enumerate(zip(true_locs, floor_ids_out)):
+    for i, (true_loc, fid) in enumerate(zip(true_locs, floor_ids_out, strict=True)):
         # Find k nearest RPs for interpolation
         if floor_id is not None:
             dists = np.linalg.norm(rp_locs - true_loc, axis=1)
@@ -171,7 +171,7 @@ def evaluate_positioning_method(method_name, method_fn, queries, true_locs, **kw
     errors = []
     times = []
 
-    for query, true_loc in zip(queries, true_locs):
+    for query, true_loc in zip(queries, true_locs, strict=True):
         t_start = time.perf_counter()
         est_loc = method_fn(query, **kwargs)
         t_end = time.perf_counter()
diff --git a/ch5_fingerprinting/example_probabilistic.py b/ch5_fingerprinting/example_probabilistic.py
index 0feec44..2e2cd51 100644
--- a/ch5_fingerprinting/example_probabilistic.py
+++ b/ch5_fingerprinting/example_probabilistic.py
@@ -65,7 +65,7 @@ def generate_test_queries(db, n_queries=100, floor_id=None, noise_std=0.0, seed=
 
     query_fingerprints = []
 
-    for true_loc, fid in zip(true_locs, floor_ids_out):
+    for true_loc, fid in zip(true_locs, floor_ids_out, strict=True):
         if floor_id is not None:
             dists = np.linalg.norm(rp_locs - true_loc, axis=1)
         else:
@@ -99,7 +99,7 @@ def evaluate_method(method_name, method_fn, queries, true_locs, **kwargs):
     errors = []
     times = []
 
-    for query, true_loc in zip(queries, true_locs):
+    for query, true_loc in zip(queries, true_locs, strict=True):
         t_start = time.perf_counter()
         est_loc = method_fn(query, **kwargs)
         t_end = time.perf_counter()
diff --git a/ch5_fingerprinting/figs/comparison_all_methods.pdf b/ch5_fingerprinting/figs/comparison_all_methods.pdf
index 7bf550e..84fdd56 100644
Binary files a/ch5_fingerprinting/figs/comparison_all_methods.pdf and b/ch5_fingerprinting/figs/comparison_all_methods.pdf differ
diff --git a/ch5_fingerprinting/figs/comparison_all_methods.png b/ch5_fingerprinting/figs/comparison_all_methods.png
index d20edd4..3ae673e 100644
Binary files a/ch5_fingerprinting/figs/comparison_all_methods.png and b/ch5_fingerprinting/figs/comparison_all_methods.png differ
diff --git a/ch5_fingerprinting/figs/comparison_all_methods.svg b/ch5_fingerprinting/figs/comparison_all_methods.svg
index d609ba3..4902699 100644
--- a/ch5_fingerprinting/figs/comparison_all_methods.svg
+++ b/ch5_fingerprinting/figs/comparison_all_methods.svg
@@ -4176,7 +4176,7 @@ L 373.539534 421.360488
 L 345.418162 421.360488 
 L 345.418162 464.100593 
 z
-" clip-path="url(#p945a551099)" style="fill: #1f77b4; stroke: #000000; stroke-linejoin: miter"/>
+" clip-path="url(#p945a551099)" style="fill: #dda0dd; stroke: #000000; stroke-linejoin: miter"/>
    </g>
    <g id="line2d_132">
     <path d="M 359.478848 464.100593 
diff --git a/ch7_slam/example_pose_graph_slam.py b/ch7_slam/example_pose_graph_slam.py
index ba46381..9633946 100644
--- a/ch7_slam/example_pose_graph_slam.py
+++ b/ch7_slam/example_pose_graph_slam.py
@@ -1025,7 +1025,7 @@ def build_map_from_poses(
     from core.slam import se2_apply
 
     all_points = []
-    for pose, scan in zip(poses, scans):
+    for pose, scan in zip(poses, scans, strict=True):
         if len(scan) > 0:
             # Transform scan to global frame
             transformed = se2_apply(pose, scan)
diff --git a/ch7_slam/example_scan_matching_visualization.py b/ch7_slam/example_scan_matching_visualization.py
index b98ae92..6e169bb 100644
--- a/ch7_slam/example_scan_matching_visualization.py
+++ b/ch7_slam/example_scan_matching_visualization.py
@@ -194,7 +194,9 @@ def plot_icp_correspondences(max_pairs: int = 45) -> plt.Figure:
             label="source scan",
         )
         stride = max(1, len(matched_source) // max_pairs)
-        for src_pt, tgt_pt in zip(matched_source[::stride], matched_target[::stride]):
+        for src_pt, tgt_pt in zip(
+            matched_source[::stride], matched_target[::stride], strict=True
+        ):
             ax.plot(
                 [src_pt[0], tgt_pt[0]],
                 [src_pt[1], tgt_pt[1]],
@@ -609,7 +611,9 @@ def update(frame: int):
             label="source scan",
         )
         stride = max(1, len(matched_source) // max_pairs)
-        for src_pt, tgt_pt in zip(matched_source[::stride], matched_target[::stride]):
+        for src_pt, tgt_pt in zip(
+            matched_source[::stride], matched_target[::stride], strict=True
+        ):
             axes[0].plot(
                 [src_pt[0], tgt_pt[0]],
                 [src_pt[1], tgt_pt[1]],
diff --git a/ch8_sensor_fusion/example_comparison.py b/ch8_sensor_fusion/example_comparison.py
index a3f10db..ba4f7be 100644
--- a/ch8_sensor_fusion/example_comparison.py
+++ b/ch8_sensor_fusion/example_comparison.py
@@ -617,7 +617,7 @@ def interpolate_truth(t_est):
     ax9.grid(True, alpha=0.3, axis="y")
 
     # Add value labels on bars
-    for i, (lc_val, tc_val) in enumerate(zip(lc_values, tc_values)):
+    for i, (lc_val, tc_val) in enumerate(zip(lc_values, tc_values, strict=True)):
         ax9.text(
             i - width / 2, lc_val, f"{lc_val:.1f}", ha="center", va="bottom", fontsize=8
         )
diff --git a/ch8_sensor_fusion/example_observability.py b/ch8_sensor_fusion/example_observability.py
index 59951b3..cd81e39 100644
--- a/ch8_sensor_fusion/example_observability.py
+++ b/ch8_sensor_fusion/example_observability.py
@@ -995,7 +995,8 @@ def main():
             # numpy 2 reprs as np.float64(-0.9999999999999998) -- accurate and
             # unreadable in a table the chapter asks the reader to interpret.
             components = {
-                name: round(float(value), 3) for name, value in zip(state_names, mode)
+                name: round(float(value), 3)
+                for name, value in zip(state_names, mode, strict=True)
             }
             print(f"    Mode {i+1}: {components}")
             # Identify dominant components. Taking argsort()[-2:] unconditionally
@@ -1039,7 +1040,7 @@ def main():
         print("\n  Unobservable modes:")
         for i in range(obs_analysis_fixes["n_unobservable"]):
             mode = obs_analysis_fixes["unobservable_modes"][:, i]
-            print(f"    Mode {i+1}: {dict(zip(state_names, mode))}")
+            print(f"    Mode {i+1}: {dict(zip(state_names, mode, strict=True))}")
     else:
         print("\n  System is FULLY OBSERVABLE!")
 
diff --git a/ch8_sensor_fusion/example_robust_tuning.py b/ch8_sensor_fusion/example_robust_tuning.py
index 11f93a9..7449537 100644
--- a/ch8_sensor_fusion/example_robust_tuning.py
+++ b/ch8_sensor_fusion/example_robust_tuning.py
@@ -613,7 +613,7 @@ def get_errors(history):
     ax7.grid(True, alpha=0.3, axis="y")
 
     # Add value labels
-    for bar, rmse in zip(bars, rmses):
+    for bar, rmse in zip(bars, rmses, strict=True):
         height = bar.get_height()
         ax7.text(
             bar.get_x() + bar.get_width() / 2.0,
@@ -648,7 +648,7 @@ def get_errors(history):
     ax8.grid(True, alpha=0.3, axis="y")
 
     # Add value labels
-    for bar, rate in zip(bars, acceptance_rates):
+    for bar, rate in zip(bars, acceptance_rates, strict=True):
         height = bar.get_height()
         ax8.text(
             bar.get_x() + bar.get_width() / 2.0,
diff --git a/ch8_sensor_fusion/example_temporal_calibration.py b/ch8_sensor_fusion/example_temporal_calibration.py
index 33f6b1c..221eaca 100644
--- a/ch8_sensor_fusion/example_temporal_calibration.py
+++ b/ch8_sensor_fusion/example_temporal_calibration.py
@@ -590,7 +590,7 @@ def get_errors(history):
     ax6.grid(True, alpha=0.3, axis="y")
 
     # Add value labels
-    for bar, val in zip(bars1, no_corr_vals):
+    for bar, val in zip(bars1, no_corr_vals, strict=True):
         height = bar.get_height()
         ax6.text(
             bar.get_x() + bar.get_width() / 2.0,
@@ -600,7 +600,7 @@ def get_errors(history):
             va="bottom",
             fontsize=8,
         )
-    for bar, val in zip(bars2, with_corr_vals):
+    for bar, val in zip(bars2, with_corr_vals, strict=True):
         height = bar.get_height()
         ax6.text(
             bar.get_x() + bar.get_width() / 2.0,
diff --git a/core/fingerprinting/classification.py b/core/fingerprinting/classification.py
index f00c93e..bfc3791 100644
--- a/core/fingerprinting/classification.py
+++ b/core/fingerprinting/classification.py
@@ -100,7 +100,7 @@ def predict(
         if return_proba and hasattr(self.classifier, "predict_proba"):
             probas = self.classifier.predict_proba(z_2d)[0]
             classes = self.classifier.classes_
-            info["class_probabilities"] = dict(zip(classes, probas))
+            info["class_probabilities"] = dict(zip(classes, probas, strict=True))
             info["top_k_classes"] = classes[np.argsort(probas)[::-1][:5]]
 
         return predicted_location, info
diff --git a/notebooks/ch3_state_estimation.ipynb b/notebooks/ch3_state_estimation.ipynb
index 1c85f56..41e243c 100644
--- a/notebooks/ch3_state_estimation.ipynb
+++ b/notebooks/ch3_state_estimation.ipynb
@@ -208,7 +208,7 @@
         "ax.plot(x_true[0], x_true[1], 'g*', markersize=20, label='True Position')\n",
         "ax.plot(x_wls[0], x_wls[1], 'r^', markersize=15, label=f'WLS Estimate (err={error_wls:.3f}m)')\n",
         "\n",
-        "for i, (anchor, r) in enumerate(zip(anchors, ranges_meas)):\n",
+        "for i, (anchor, r) in enumerate(zip(anchors, ranges_meas, strict=True)):\n",
         "    circle = plt.Circle(anchor, r, fill=False, linestyle='--', alpha=0.5)\n",
         "    ax.add_patch(circle)\n",
         "\n",
diff --git a/notebooks/ch4_rf_positioning.ipynb b/notebooks/ch4_rf_positioning.ipynb
index ca7d3d8..d269da4 100644
--- a/notebooks/ch4_rf_positioning.ipynb
+++ b/notebooks/ch4_rf_positioning.ipynb
@@ -237,7 +237,7 @@
         "true_ranges = np.array([toa_range(anchor, true_pos) for anchor in anchors])\n",
         "\n",
         "print(f\"\\n📏 True ranges to each anchor:\")\n",
-        "for i, (anchor, r) in enumerate(zip(anchors, true_ranges)):\n",
+        "for i, (anchor, r) in enumerate(zip(anchors, true_ranges, strict=True)):\n",
         "    print(f\"  A{i+1}: {r:.4f} m\")\n",
         "\n",
         "# Solve with 1/d^2 range weighting ('iwls' was a deprecated alias for this)\n",
@@ -273,7 +273,7 @@
         "noisy_ranges = true_ranges + np.random.randn(len(anchors)) * noise_std\n",
         "\n",
         "print(f\"\\n📏 Range measurements (σ = {noise_std*100:.0f} cm):\")\n",
-        "for i, (true_r, noisy_r) in enumerate(zip(true_ranges, noisy_ranges)):\n",
+        "for i, (true_r, noisy_r) in enumerate(zip(true_ranges, noisy_ranges, strict=True)):\n",
         "    error_r = noisy_r - true_r\n",
         "    print(f\"  A{i+1}: True={true_r:.4f}m, Measured={noisy_r:.4f}m, Error={error_r*100:+.1f}cm\")\n",
         "\n",
@@ -306,7 +306,7 @@
         "            fontsize=11, fontweight='bold')\n",
         "\n",
         "# Plot range circles (true ranges)\n",
-        "for anchor, r in zip(anchors, true_ranges):\n",
+        "for anchor, r in zip(anchors, true_ranges, strict=True):\n",
         "    circle = Circle(anchor, r, fill=False, color='red', alpha=0.3, \n",
         "                    linestyle='--', linewidth=1.5)\n",
         "    ax.add_patch(circle)\n",
@@ -389,7 +389,7 @@
         "tdoa_noisy = tdoa_measurements + np.random.randn(len(tdoa_measurements)) * tdoa_noise_std\n",
         "\n",
         "print(f\"\\n📏 Noisy TDOA measurements (σ = {tdoa_noise_std*100:.0f} cm):\")\n",
-        "for i, (true_t, noisy_t) in enumerate(zip(tdoa_measurements, tdoa_noisy), start=1):\n",
+        "for i, (true_t, noisy_t) in enumerate(zip(tdoa_measurements, tdoa_noisy, strict=True), start=1):\n",
         "    print(f\"  Δr_{i+1},{reference_idx+1} = {noisy_t:+.4f} m (error: {(noisy_t-true_t)*100:+.1f} cm)\")\n",
         "\n",
         "# Solve using TDOA positioner\n",
@@ -445,7 +445,7 @@
         "true_aoa = np.array([aoa_azimuth(anchor, true_pos) for anchor in anchors])\n",
         "\n",
         "print(f\"\\n📐 True AOA measurements (azimuth angles):\")\n",
-        "for i, (anchor, theta) in enumerate(zip(anchors, true_aoa)):\n",
+        "for i, (anchor, theta) in enumerate(zip(anchors, true_aoa, strict=True)):\n",
         "    print(f\"  A{i+1}: θ = {np.rad2deg(theta):+.2f}°\")\n",
         "\n",
         "# Add noise\n",
@@ -455,7 +455,7 @@
         "aoa_noisy = true_aoa + np.random.randn(len(anchors)) * aoa_noise_std_rad\n",
         "\n",
         "print(f\"\\n📐 Noisy AOA measurements (σ = {aoa_noise_std_deg:.0f}°):\")\n",
-        "for i, (true_a, noisy_a) in enumerate(zip(true_aoa, aoa_noisy)):\n",
+        "for i, (true_a, noisy_a) in enumerate(zip(true_aoa, aoa_noisy, strict=True)):\n",
         "    error_deg = np.rad2deg(noisy_a - true_a)\n",
         "    print(f\"  A{i+1}: θ = {np.rad2deg(noisy_a):+.2f}° (error: {error_deg:+.2f}°)\")\n",
         "\n",
@@ -534,7 +534,7 @@
         "])\n",
         "\n",
         "print(f\"\\n📏 Ranges estimated from RSS:\")\n",
-        "for i, (true_r, est_r) in enumerate(zip(true_ranges, ranges_from_rss)):\n",
+        "for i, (true_r, est_r) in enumerate(zip(true_ranges, ranges_from_rss, strict=True)):\n",
         "    error_pct = 100 * (est_r - true_r) / true_r\n",
         "    print(f\"  A{i+1}: True={true_r:.2f}m, Estimated={est_r:.2f}m (error: {error_pct:+.1f}%)\")\n",
         "\n",
@@ -634,7 +634,7 @@
         "\n",
         "ax1 = axes[0]\n",
         "bp = ax1.boxplot(solved_errors, tick_labels=methods, patch_artist=True)\n",
-        "for patch, color in zip(bp['boxes'], colors):\n",
+        "for patch, color in zip(bp['boxes'], colors, strict=True):\n",
         "    patch.set_facecolor(color)\n",
         "    patch.set_alpha(0.7)\n",
         "ax1.set_ylabel('Position error (m)', fontsize=12)\n",
@@ -642,7 +642,7 @@
         "ax1.grid(True, alpha=0.3, axis='y')\n",
         "\n",
         "ax2 = axes[1]\n",
-        "for method, color in zip(methods, colors):\n",
+        "for method, color in zip(methods, colors, strict=True):\n",
         "    errors = np.sort(outcomes[method].errors[outcomes[method].solved])\n",
         "    cdf = np.arange(1, len(errors) + 1) / len(errors)\n",
         "    ax2.plot(errors, cdf, label=method, color=color, linewidth=2)\n",
@@ -659,7 +659,7 @@
         "ax3 = axes[2]\n",
         "failed = [100 * outcomes[m].n_failed / outcomes[m].n for m in methods]\n",
         "bars = ax3.bar(methods, failed, color=colors, alpha=0.7, edgecolor='black')\n",
-        "for bar, m in zip(bars, methods):\n",
+        "for bar, m in zip(bars, methods, strict=True):\n",
         "    ax3.text(bar.get_x() + bar.get_width() / 2, bar.get_height() + 2,\n",
         "             f\"{outcomes[m].n_failed}/{outcomes[m].n}\", ha='center', fontsize=11)\n",
         "ax3.set_ylabel('Fixes that failed (%)', fontsize=12)\n",
diff --git a/notebooks/ch5_fingerprinting.ipynb b/notebooks/ch5_fingerprinting.ipynb
index 88367de..5b0975f 100644
--- a/notebooks/ch5_fingerprinting.ipynb
+++ b/notebooks/ch5_fingerprinting.ipynb
@@ -310,7 +310,7 @@
         "nn_errors = []\n",
         "nn_estimates = []\n",
         "\n",
-        "for query, true_loc in zip(queries, true_locs):\n",
+        "for query, true_loc in zip(queries, true_locs, strict=True):\n",
         "    est_loc = nn_localize(query, db, metric=\"euclidean\", floor_id=floor_id)\n",
         "    nn_estimates.append(est_loc)\n",
         "    nn_errors.append(np.linalg.norm(est_loc - true_loc))\n",
@@ -344,7 +344,7 @@
         "    knn_errors = []\n",
         "    knn_estimates = []\n",
         "    \n",
-        "    for query, true_loc in zip(queries, true_locs):\n",
+        "    for query, true_loc in zip(queries, true_locs, strict=True):\n",
         "        est_loc = knn_localize(query, db, k=k, metric=\"euclidean\", \n",
         "                               weighting=\"inverse_distance\", floor_id=floor_id)\n",
         "        knn_estimates.append(est_loc)\n",
@@ -415,7 +415,7 @@
         "# MAP positioning\n",
         "print(\"\\n🎯 MAP Positioning (Eq. 5.4):\")\n",
         "map_errors = []\n",
-        "for query, true_loc in zip(queries, true_locs):\n",
+        "for query, true_loc in zip(queries, true_locs, strict=True):\n",
         "    est_loc = map_localize(query, model, floor_id=floor_id)\n",
         "    map_errors.append(np.linalg.norm(est_loc - true_loc))\n",
         "map_errors = np.array(map_errors)\n",
@@ -426,7 +426,7 @@
         "# Posterior Mean positioning\n",
         "print(\"\\n🎯 Posterior Mean Positioning (Eq. 5.5):\")\n",
         "pm_errors = []\n",
-        "for query, true_loc in zip(queries, true_locs):\n",
+        "for query, true_loc in zip(queries, true_locs, strict=True):\n",
         "    est_loc = posterior_mean_localize(query, model, floor_id=floor_id)\n",
         "    pm_errors.append(np.linalg.norm(est_loc - true_loc))\n",
         "pm_errors = np.array(pm_errors)\n",
@@ -484,7 +484,7 @@
         "colors = ['#3498db', '#e74c3c', '#2ecc71', '#9b59b6', '#f39c12']\n",
         "bp = ax1.boxplot([all_results[m] for m in all_results], \n",
         "                  tick_labels=list(all_results.keys()), patch_artist=True)\n",
-        "for patch, color in zip(bp['boxes'], colors):\n",
+        "for patch, color in zip(bp['boxes'], colors, strict=True):\n",
         "    patch.set_facecolor(color)\n",
         "    patch.set_alpha(0.7)\n",
         "ax1.set_ylabel('Position Error (m)', fontsize=12)\n",
@@ -494,7 +494,7 @@
         "\n",
         "# Plot 2: CDF comparison\n",
         "ax2 = axes[1]\n",
-        "for (method, errors), color in zip(all_results.items(), colors):\n",
+        "for (method, errors), color in zip(all_results.items(), colors, strict=True):\n",
         "    sorted_errors = np.sort(errors)\n",
         "    cdf = np.arange(1, len(sorted_errors) + 1) / len(sorted_errors)\n",
         "    ax2.plot(sorted_errors, cdf, label=method, color=color, linewidth=2)\n",
diff --git a/scripts/generate_ch2_coordinate_transforms_dataset.py b/scripts/generate_ch2_coordinate_transforms_dataset.py
index 1eba2b1..7287a7b 100644
--- a/scripts/generate_ch2_coordinate_transforms_dataset.py
+++ b/scripts/generate_ch2_coordinate_transforms_dataset.py
@@ -85,7 +85,10 @@ def generate_building_trajectory_llh(
     north_m = rng.uniform(-half, half, n_points)
     east_m = rng.uniform(-half, half, n_points)
     offsets = np.array(
-        [enu_to_llh_offset(e, n, lat_center) for e, n in zip(east_m, north_m)]
+        [
+            enu_to_llh_offset(e, n, lat_center)
+            for e, n in zip(east_m, north_m, strict=True)
+        ]
     )
 
     # Generate random positions within building footprint
@@ -274,7 +277,10 @@ def generate_dataset(
     # Convert to ECEF
     print("\nStep 2: Converting LLH -> ECEF...")
     ecef = np.array(
-        [llh_to_ecef(lat, lon, h) for lat, lon, h in zip(lats, lons, heights)]
+        [
+            llh_to_ecef(lat, lon, h)
+            for lat, lon, h in zip(lats, lons, heights, strict=True)
+        ]
     )
     print(
         f"  ECEF X range: {ecef[:, 0].min()/1e3:.1f}km to {ecef[:, 0].max()/1e3:.1f}km"
diff --git a/scripts/generate_ch6_env_sensors_dataset.py b/scripts/generate_ch6_env_sensors_dataset.py
index 655b842..85dd650 100644
--- a/scripts/generate_ch6_env_sensors_dataset.py
+++ b/scripts/generate_ch6_env_sensors_dataset.py
@@ -567,7 +567,12 @@ def generate_dataset(
     # and got it wrong -- duplicated policy only has to be forgotten once.
     heading_true = att_true[:, 2]
     heading_error = np.abs(
-        np.array([wrap_angle_diff(e, t_) for e, t_ in zip(heading_est, heading_true)])
+        np.array(
+            [
+                wrap_angle_diff(e, t_)
+                for e, t_ in zip(heading_est, heading_true, strict=True)
+            ]
+        )
     )
     heading_error_deg = np.rad2deg(heading_error)
     mean_heading_error = np.mean(heading_error_deg)
diff --git a/tests/ch6_dead_reckoning/test_comparison_figures.py b/tests/ch6_dead_reckoning/test_comparison_figures.py
index ad02f67..4146a4a 100644
--- a/tests/ch6_dead_reckoning/test_comparison_figures.py
+++ b/tests/ch6_dead_reckoning/test_comparison_figures.py
@@ -70,7 +70,9 @@ def test_same_seed_gives_identical_measurements(self):
         first = self._noise(DEFAULT_SEED)
         second = self._noise(DEFAULT_SEED)
 
-        for label, a, b in zip(("accel", "gyro", "mag", "wheel"), first, second):
+        for label, a, b in zip(
+            ("accel", "gyro", "mag", "wheel"), first, second, strict=True
+        ):
             with self.subTest(sensor=label):
                 np.testing.assert_array_equal(a, b)
 
diff --git a/tests/ch6_dead_reckoning/test_heading_error_is_wrapped.py b/tests/ch6_dead_reckoning/test_heading_error_is_wrapped.py
index 67da326..203c3f1 100644
--- a/tests/ch6_dead_reckoning/test_heading_error_is_wrapped.py
+++ b/tests/ch6_dead_reckoning/test_heading_error_is_wrapped.py
@@ -67,7 +67,10 @@ def setUpClass(cls):
         )
         cls.error = np.abs(
             np.array(
-                [wrap_angle_diff(e, y) for e, y in zip(cls.heading_est, cls.yaw_true)]
+                [
+                    wrap_angle_diff(e, y)
+                    for e, y in zip(cls.heading_est, cls.yaw_true, strict=True)
+                ]
             )
         )
 
diff --git a/tests/ch7_slam/test_slam_frontend_figure.py b/tests/ch7_slam/test_slam_frontend_figure.py
index f6dc774..e7b72f8 100644
--- a/tests/ch7_slam/test_slam_frontend_figure.py
+++ b/tests/ch7_slam/test_slam_frontend_figure.py
@@ -232,7 +232,7 @@ def test_writing_the_figure_twice_is_byte_identical(self):
                 first = save_figure(fig, Path(tmp) / "a", FIGURE_NAME)
                 second = save_figure(fig, Path(tmp) / "b", FIGURE_NAME)
 
-                for lhs, rhs in zip(first, second):
+                for lhs, rhs in zip(first, second, strict=True):
                     with self.subTest(fmt=lhs.suffix):
                         self.assertEqual(lhs.read_bytes(), rhs.read_bytes())
         finally:
diff --git a/tests/core/eval/test_plots.py b/tests/core/eval/test_plots.py
index 6318c8e..516e15e 100644
--- a/tests/core/eval/test_plots.py
+++ b/tests/core/eval/test_plots.py
@@ -216,7 +216,7 @@ def test_output_is_byte_reproducible(self, tmp_path):
         finally:
             plt.close(fig)
 
-        for lhs, rhs in zip(first, second):
+        for lhs, rhs in zip(first, second, strict=True):
             assert (
                 lhs.read_bytes() == rhs.read_bytes()
             ), f"{lhs.suffix} output is not reproducible"
diff --git a/tests/core/fusion/test_fusion_tuning.py b/tests/core/fusion/test_fusion_tuning.py
index 25578af..953a435 100644
--- a/tests/core/fusion/test_fusion_tuning.py
+++ b/tests/core/fusion/test_fusion_tuning.py
@@ -9,6 +9,7 @@
 
 import unittest
 import warnings
+from itertools import pairwise
 
 import numpy as np
 
@@ -527,7 +528,9 @@ def test_scale_grows_monotonically_with_the_residual(self):
         cauchy_scales = [cauchy_R_scale(r, 2.385) for r in self.RESIDUALS]
 
         for scales in (huber_scales, cauchy_scales):
-            for earlier, later in zip(scales, scales[1:]):
+            # pairwise, not zip(scales, scales[1:]): that idiom relies on
+            # zip truncating, so strict=True raises on every call.
+            for earlier, later in pairwise(scales):
                 self.assertLessEqual(earlier, later + 1e-12)
 
     def test_literal_printed_form_would_trust_outliers_more(self):
diff --git a/tests/core/slam/test_submap_2d.py b/tests/core/slam/test_submap_2d.py
index e7ecec8..c6845ba 100644
--- a/tests/core/slam/test_submap_2d.py
+++ b/tests/core/slam/test_submap_2d.py
@@ -277,7 +277,7 @@ def test_build_submap_from_trajectory(self):
             np.array([[3.0, 0.0]]),  # From pose 2
         ]
 
-        for pose, scan in zip(poses, scans):
+        for pose, scan in zip(poses, scans, strict=True):
             submap.add_scan(pose, scan)
 
         self.assertEqual(submap.n_scans, 3)
diff --git a/tests/docs/test_readme_example_output.py b/tests/docs/test_readme_example_output.py
index ce35b1f..1fe1319 100644
--- a/tests/docs/test_readme_example_output.py
+++ b/tests/docs/test_readme_example_output.py
@@ -172,7 +172,7 @@ def _matches(expected: str, actual: str) -> bool:
     got = actual.split(" ")
     if len(want) != len(got):
         return False
-    return all(_token_matches(w, g) for w, g in zip(want, got))
+    return all(_token_matches(w, g) for w, g in zip(want, got, strict=True))
 
 
 def _find_from(live, expected, start):
diff --git a/tests/test_lint_debt_only_shrinks.py b/tests/test_lint_debt_only_shrinks.py
index c0bab41..11dc000 100644
--- a/tests/test_lint_debt_only_shrinks.py
+++ b/tests/test_lint_debt_only_shrinks.py
@@ -30,9 +30,9 @@
 **What is left is mostly not lint at all.** 727 of the remaining findings are
 UP006/UP045/UP035/UP007 -- `List[int]` for `list[int]`, `Optional[X]` for
 `X | None`. Those became legal only when the floor moved to 3.10, they are
-mechanical, and they are worth doing in their own change. The ~140 after that
-are the ones with actual content: 41 `zip()` calls without `strict=`, which
-truncate silently to the shorter argument, are the interesting ones.
+mechanical, and they are worth doing in their own change. The ~100 after that
+are the ones with actual content -- B905 was the largest of them and is now
+gone, audited rather than swept: see the comment on BASELINE below.
 
 Per-rule rather than a single total on purpose: a total lets ten fixed W293
 pay for ten new B905, which is the opposite of what a ratchet is for.
@@ -65,9 +65,16 @@
 #: Findings per rule today. Only ever edit these downwards.
 #:
 #: UP0xx are the annotation modernisations that the 3.10 floor made available,
-#: and are the bulk of what is left. B905 is the one group worth reading before
-#: fixing: `zip()` without `strict=` truncates to the shorter argument without
-#: saying so. The twelve remaining W293 sit inside argparse `epilog` strings,
+#: and are the bulk of what is left.
+#:
+#: B905 is absent, and its absence is the guard: the `appeared` check below
+#: fails on any rule not listed here, so one new `zip()` without `strict=`
+#: turns this red.
+#: The 41 it used to record were read one at a time rather than swept, which is
+#: how the two that mattered were found -- a boxplot palette one shade short of
+#: its methods, and one site where `strict=True` is simply wrong, because
+#: `zip(xs, xs[1:])` relies on the truncation. The twelve remaining W293 sit
+#: inside argparse `epilog` strings,
 #: where the whitespace is content that gets printed rather than layout --
 #: black leaves those alone, correctly, and so should you.
 BASELINE = {
@@ -75,7 +82,6 @@
     "UP045": 178,
     "UP035": 123,
     "I001": 84,
-    "B905": 41,
     "UP007": 38,
     "B007": 28,
     "B028": 14,