The gate reports this itself:
```
[check-inert-flag-seams] STALE allow-list entries — supplied now, or no longer declared:
plugins/fusion-plugin-dependency-graph/src/GraphTaskNode.tsx::isTaskStuck
— no unsupplied call site remains; remove its ALLOWED_OMISSIONS entry
```
#3029 wired that call site. The entry now claims to tolerate something
that doesn't exist.
The gate **warns rather than fails** here by design — failing would
punish whoever fixed the seam — which is precisely why a stale entry has
to be deleted deliberately instead of waiting for a red build to force
it.
## The text is worth losing on its own account
That exemption justified the omission as needing *"a published-API
change."* I later revised it to *"build plumbing."* **Both were wrong**:
the blocker was a hand-maintained interop declaration frozen at the
pre-conversion three-argument signature (#3003 → #3029).
An exemption whose stated reason has been disproven twice is worse than
no exemption — it's a confident note that stops anyone re-deriving the
answer. Same failure mode as the stale "do not re-probe" doc entry
corrected in #3018, and the untested deferral rationales recorded in
#3026.
## Measured
| check | result |
|---|---|
| inert-seam gate | **23 seams, all supplied**, exit 0 **with no
waiver** |
| gate's own suite | 18/18 |
| lane-wiring · plugin-interop-drift · FNXC | green; lint clean |
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **Bug Fixes**
* Improved validation of task status handling by removing an exception
that could allow an incomplete call site to go undetected.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
6a5bd7a86f