#299 ruling taken, no second run, no remedy started — waiting on your relay of the operator pick. I will hold the 457 ms pre-subnet block and the 30 s-dial-vs-10 s-ceiling disagreement for the #299 closing comment rather than posting now, since the issue does not close until the remedy lands. Say if you want them posted as an interim comment instead.

#49/#267, two corrections done.

1. Q3 wording, per your note. Was: "the DRAINED gate read NOT drained on its first evaluation after ready". That reads as a global ordinal. Now: "the DRAINED gate read NOT drained; this is the first such observation in THIS trial, not the first evaluation of the gate; the trial keeps polling to its window". The doc comment above it gained the same scoping — earlier trials evaluated the same gate and this line says nothing about them. Corrected by replacement, no annotation.

2. CLOSE_WRITER_GONE rename — and the defect is worse than a word. Its ONE call site is `mark_controller_gone`, the W1 deadline-evict path: the drain timed out the controller's `send_timeout` and the epoch guard still matched. A writer BLOCKED past its deadline has not been shown to have exited or dropped its `tx` — which is exactly what "writer_gone" asserts. The name claimed a fate the observation never established, on the only path that emits it.
   Renamed to `CLOSE_SEND_DEADLINE` / value `send_deadline` — it names the deadline the drain observed and nothing about the writer. Doc comment rewritten to say so explicitly, and the plan doc's edge table sentence (line 129, "a writer that exited (`writer_gone`)") corrected by replacement to match. I picked the word myself; if you want a different one, say it and it is a one-line change.
   Note the two remaining causes are clean: `stall_evict` and `detach` both name their caller's knowledge.

Check status: the FIRST `cargo check` I started overlapped my brainproc.rs edit, so it read a torn tree — I discarded it rather than count it, stopped the task, swept for survivor cargo/rustc (post-state clean, none found) and started a clean `cargo check -p spt-daemon --jobs 2` on the final tree. That is running now. `traceable-reqs check` at the pinned CI version follows it. I will report both results with the check-green, not before.

#300 amended scope read; not started, queued behind the check.