compiler-output-regression: native_owned_typed_views misses vectorization for an unexpected control_flow reason (Linux-only, pre-existing, blocking full-suite-gate)

[proggeramlug]

compiler-output-regression fails its "Gate native-ABI proof compiler output" step on one workload:

native_owned_typed_views
vectorization_expectation:
  observed_missed_reason_kinds: {"call_instruction": 4, "control_flow": 3,
                                 "generic_not_vectorized": 7, "unknown_trip_count": 1}
  unexpected_missed_reason_kinds: ["control_flow"]
  min_vectorized_loops: 0, vectorized_count: 0, passed: false

control_flow is not in that workload's allowed_missed_reason_kinds, so three loops now miss vectorization for a reason the gate has not seen before.

Scope / attribution

• Pre-existing, and not from the recent merge batch. It fails identically on the attribution baseline 3627657c7 (run 32368807258), which contains none of fix(hir): ambient declare function require must not shadow the require intrinsic #8452/fix(link): dedup evidence must match what the linker can actually resolve (#8455) #8458/test(parity): ratchet the matrix-trend baseline over the #8271 dark debt #8461/ci: give full-tier jobs timeout margin over observed contention #8462/fix(runtime): extend #8416's C-unwind to the whole closure dispatch family (#8463) #8464, and it still fails after revert: back out #8464's blanket C-unwind conversion (measured +20 crashes, gc-stress) #8484 reverted fix(runtime): extend #8416's C-unwind to the whole closure dispatch family (#8463) #8464 (run 32377162076). So it is not caused by the extern "C-unwind" unwind edges, which was the obvious first suspect.
• Not reproducible on macOS. Running scripts/compiler_output_regression.py suite --suite native-abi-proof --gate locally against a perry-dev build reports failed_workloads: []. This appears to be specific to the Linux/release configuration the job uses.
• The likeliest origin is the recent string-codegen work (refactor(stdlib): centralize string header readers #8448 / perf(codegen): fuse String.concat arguments into concat chain #8450 / fix(ui): replace borrowed runtime strings with owned copies #8453 / perf(codegen): append strings through variable cells #8454), which is the codegen change in the window, but that is not established — attributing it wants a bisect on Linux.

Why this is a low-severity gate failure: this workload sets min_vectorized_loops = 0 with scalar_baseline = "allowed: native-owned typed view fixture gates proof records and fallback reasons". It does not require vectorization; the reason list is a change detector. Sibling workloads (native_abi_packet_control, native_pod_layout_constants) already list control_flow as acceptable.

It is, however, blocking the release: compiler-output-regression is one of full-suite-gate's required jobs.

Proposed: widen this workload's allowed_missed_reason_kinds to include control_flow, matching its siblings, with a comment pointing here — and keep this issue open to attribute the codegen change that introduced it. That trades a precise change-detector for an unblocked release, so it should be reverted once the cause is known and either fixed or accepted deliberately.