Resource-free long waits as the flagship (hardened pause/wait + restart-survival demo)

[mcclowes]

Phase 3 · Differentiate on the domain

Make resource-free long waits the flagship use case: wait days/weeks for a webhook or external job while consuming zero compute, then resume exactly where you left off.

Goals

• Harden pause/wait on the durable foundation: atomic resume (Paused executions can resume twice (timeout poller vs webhook race, no atomic status) #130), durable timers that survive restart and don't overflow (Long pauses overflow setTimeout (>24.8d fires instantly); timers lost on restart #131), webhook-vs-timeout race resolved.
• Target use cases: human-in-the-loop approval flows, async API callbacks, polling an external job to completion.
• A demo mission that pauses for a webhook, survives a process restart, and resumes — shown end to end.

Why

This is the visible magic Temporal charges a cluster for; for Reqon it's a declarative one-liner.

Part of the durable-execution roadmap (see the epic).

[mcclowes]

Flagship landed in #195: a mission pauses for a webhook, the process restarts, and the run resumes from the durable log and finishes. Includes the correctness fix so resume continues past the pause (restoring the checkpoint + webhook payload) instead of re-pausing, a runnable demo (examples/durable-approval/, file-backed log + separate executor for the resume to model a real restart), and a restart-survival end-to-end test. Built on the durable pause foundation from #192.