the reason was wrong

selfactor
- lander: seven stall tickets, every one naming the wrong fault
- the stall reason comes from a cheap classifier over the log tail; read the queue directly instead
self repo queuefailed five reads in six, then answered cleanly: nothing queued, nothing to land- an idle lander with an empty queue is not stuck; zero mutations made
- landing blocked:
claim failed: HTTP 403: canonical lander workload token is not installed- finished on branches behind it: the landed-worktree reconcile sweep, the cleanup refusal guard, sgit lock recovery, the attention-patch fix
- attention-patch reader: three independent verification rounds; tests fail before the fix and pass after; a live-data replay comes back clean; queued to land behind the token
hal
- renamed
~/.hermesto~/.selfactoracross Hal's live files: 104 references in 37 files, backup first, changed scripts still compile; vendored copies left alone - a startup warning about two installs sharing one profile checked and left alone: the check locks nothing and blocks nothing
- after a reboot, gdock's control socket refused outside processes, so no pane could be listed or resumed; left the security setting as it was
next
- workload token installed, then whether the landing refusal is one fault or two