#2472 serve --hot-reload: a changed trigger override schedule is logged as applied but the built-in cron keeps firing

Problem

With swamp serve --hot-reload, changing a trigger override's schedule in the --config file and sending swamp serve reload (SIGHUP) logs that the schedule changed. But the scheduler keeps firing on the workflow's built-in cron.

Measured (20260923.231117.0)

serve runs with --config=/config/serve.yaml --hot-reload. The workflow declares trigger.schedule: "0 5 * * *". The config's triggers: override for it was changed to schedule: "44 10 * * *", confirmed in the pod's mounted file, then:

$ swamp serve reload --repo-dir /data/repo
Sent SIGHUP to serve process 68

Server log, in order:

SIGHUP received, reloading pulled extensions...
Scheduled workflow "mircloud-sweep-cabling-drift" ("0 5 * * *")
Registered schedule for workflow "mircloud-sweep-cabling-drift": "0 5 * * *"
Updated trigger overrides: 1 schedule(s) changed
Hot-reloaded 65 type(s)
Reloaded 1 trigger override(s) from serve.yaml

• GET /health afterwards: {"cronExpression":"0 5 * * *","nextRun":"2026-09-25T05:00:00.000Z"}.
• No fire at 10:44. The worker received 0 dispatches for the workflow.
• Control: the same override applied by a restart (07:15Z, cron 21 07 * * *) fired on time, run e47f6fa5 at 07:21:00Z.

Suspected cause (INFERRED from the log order)

The extension hot-reload rescans workflow directories and re-registers each workflow's built-in schedule, and that overwrites the override the reload just applied. It may also discard the override's inputs. That is not measured, and if true it would make a scheduled fire fail input validation after any reload.

Ask

• Apply trigger overrides AFTER the workflow rescan on reload, or make the rescan override-aware.
• Make /health's schedules[] report the effective cron and the source (built-in or override).