[apscheduler] make the managed job store immutable at runtime

[ricardo-agz]

Rejects runtime job creation and mutation on the managed store

Rationale:

Runtime cache is not durable enough / expose the api needed to reliably implement dynamic job creation / mutation

Eviction would mean a runtime-added job would get dropped, or a modified job would silently revert back to it's original config, we should enable this later once our store can offer those durability guarantees

scheduler.pause_job("cleanup")
# APSchedulerConfigurationError: cannot update job "cleanup" at runtime:
# the managed job store is immutable at runtime; change the declaration
# and deploy, or gate the job's work with state your application owns

The gate lives in the coordinator, not the patched scheduler methods, so in-wake self-mutations (which bypass the wrappers in queue-serving runtimes) are covered by the same check.

With runtime writes gone, every record is a declaration plus progress, so the provenance concept (declared vs runtime records) is meaningless and this PR deletes it outright — nothing has shipped, so there is no legacy state to respect.

Deliberately unchanged

• Declaration-time add_job() and scheduled_job() (that is how declared jobs exist).
• Execution progress writes (cas_update/cas_remove from the wake path) and reconciliation.
• Scheduler-level start()/pause()/resume() — documented as operational control; pause() is not a kill switch.

A future durable backend can lift the restriction behind the same Backend protocol without an API change.

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
vercel-py	Ready	Preview	Aug 12, 2026 2:20am