quest: plan the moq-gst sibling-restart resume flake

quest/m1/test-flakes-2/README.md

@@ -46,5 +46,6 @@ Public API: none. Wire: none.
 - [macOS receive credit](/quest/m1/test-flakes-2/macos-receive-credit.md) - `partial_control_requests_give_way` passes on the macOS lane, after checking for a shared platform cause
 - [moq-mux missing sequence](/quest/m1/test-flakes-2/mux-missing-sequence.md) - `missing_sequence_skips_on_live_track` runs on a paused clock
 - [moq-rtsp stall timers](/quest/m1/test-flakes-2/rtsp-stall-timers.md) - the 200 ms stall and keyframe tests run on a mocked clock
+- [moq-gst sibling resume](/quest/m1/test-flakes-2/gst-sibling-resume.md) - an idle pad reads its named epoch again after a sibling restart, a likely liveness bug fixed before the test
 - [moq-gst pinned replay](/quest/m1/test-flakes-2/gst-pinned-replay.md) - the pinned-instance finish test waits on events instead of sleep loops
 - [Dart leak RSS](/quest/m1/test-flakes-2/dart-leak-rss.md) - the leak tests measure what moq owns, not process RSS

quest/m1/test-flakes-2/gst-sibling-resume.md

@@ -0,0 +1,38 @@
+# [M] moq-gst resumes a named epoch after a sibling restart
+
+## Goal
+
+`a_sibling_restart_idles_a_reference_naming_the_old_epoch`
+(`rs/moq-gst/src/source/imp.rs:4175`) never fails under load. Once the
+instance a reference names serves its path again, the source always goes
+back to reading it on the same pad.
+
+## Plan
+
+Decided 2026-10-25 (maintainer): treat it as a liveness bug first, then the
+test. It failed once under host load during a `just check` for
+[moq-dev/moq#6464](https://github.com/moq-dev/moq/pull/6464), and passed 5 of
+5 times in isolation and in the full suite. The failure was
+`3 never reached a pad: [Pad("video_0"), Segment("video_0", 46000), Buffer("video_0", 1)]`
+at `Recorder::written`: after the replacing epoch dropped and the second
+Restart arrived, the old instance wrote tag 3 every 50 ms for 10 s and none
+of it reached the pad. A delay that long points to a missed resume, not a
+slow test.
+
+- Reproduce first: under load, or deterministically by forcing the order of
+  the Restart and the old instance's re-subscription with a
+  `moq_clock::Manual` clock.
+- Find why the idle pad never reads the named instance again, and fix it in
+  moq-gst or moq-net, wherever the resume is lost.
+- Then make the test wait on the announce and pad events instead of the
+  sleep-polled `written()`.
+
+Separate from [gst-pinned-replay](/quest/m1/test-flakes-2/gst-pinned-replay.md)
+by decision: it uses the same `Recorder` fixture, but the cause is likely
+different.
+
+Public API: none expected. Wire: none.
+
+## Related
+
+- [moq-gst pinned replay](/quest/m1/test-flakes-2/gst-pinned-replay.md) - the same fixture's sleep-poll race