Known pre-input Hyprland action rejection leaves application-group observations permanently blocked

[Calmingstorm]

Severity: P2

Verified at v4.10.0, 55c15761845b00931ab2ea4aa184a7797dfd0614.

Locations

src/computer/runtime/hyprland_guardian.py:287-311,168-171; src/computer/runtime/hyprland_backend.py:1079-1081,1510-1519; assets/hyprland-input/guardian.c:852-855

Failure scenario / concrete benefit

Unsupported key/text/overbudget plan rejected before input with acknowledged cleanup and guardian alive. _group_refresh_cleanfalse set beforeact remains false onexception; nextfreshobservation refuses ownedcleanup despiteknownnoinput.

Verification

Independently traced exception-latch/readiness consumers and nativepreinputreject; reviewer actualPython/inertsend receipt reproduced zero refresh calls/cleanupunverified.

Correction and tests

Preserve bounded known-preinputcleanproof orretire/rearmwithoutclaimingarbitraryexceptionsclean.

Duplicate screening

Checked original open/closed inventory and fresh GitHub title/body searches. No matching root; related #404 specifically compared. No live desktop or historical native experiments performed.

[Calmingstorm]

Validated as fixed on campaign/v4.10.0-review (PR #629) at a978f193: the fix was reviewed against the final code, and its regression tests pass on the branch and in CI. It is not merged to master yet. Closing per the maintainer's instruction; reopen if a later change regresses it.