worker: the scheduler double-emits every recurring job if outpost-worker scales past 1

packages/outpost/queue/src/scheduler.ts:79 schedules recurring jobs with a plain in-process interval:

const timer = setInterval(() => this.tick(definition), definition.intervalMs);

apps/worker/src/index.ts:94 constructs one Scheduler per process and starts it at :117. With a single outpost-worker replica this is correct. At two replicas, both tick, and every scheduled job type — SLA_CHECK, ONBOARDING_DIGEST, GITHUB_REACTION_POLL, ACCOUNT_SCORING, JOB_CLEANUP — is enqueued twice per interval.

Why it matters

It is a scaling landmine rather than a live defect: nothing warns you, and the first symptom would be duplicate outbound posts from the digest, which is exactly the class of bug #133 was filed to prevent.

Fix

Leader election on the SystemConfig table — a lease row the scheduler must hold, renewed each tick, so exactly one replica emits. Alternatively, split the scheduler into its own single-replica service. Either way, worth landing before anyone scales the worker, not after.

Start with packages/outpost/queue/src/scheduler.ts:79 and apps/worker/src/index.ts:94,117 to trace scheduler construction and startup. Inspect the SystemConfig table and existing issue #133 for relevant coordination context. Done means recurring jobs are emitted by only one worker replica when outpost-worker scales past one, without duplicate enqueues.