Interrupt and Resume — References

Durable execution, process-level checkpointing, and the snapshot problem as it was originally stated.

Durable continuation

• Temporal documentation — event-history replay, durable timers, external signals, and the versioning problem for executions paused across a code change.
• AWS Step Functions documentation — callback tokens for waiting on external systems or people with no compute held, including token timeouts and heartbeats.

Capturing a running process

• CRIU — Checkpoint/Restore In Userspace — freezing and restoring a live process tree, and the documented limits around open connections and external resources that make this insufficient on its own for agent runs.
• POSIX.1 base specifications — signal semantics and the long-established distinction between requesting a stop and forcing one, which is the same distinction safe stopping points formalise.

Snapshots of distributed work

• Distributed Snapshots: Determining Global States of Distributed Systems (Chandy and Lamport, 1985) — why a consistent snapshot of a system with messages in flight requires recording channel contents, which is the formal statement of the in-flight-call problem. Cited by name.

Not repeating an applied effect

• RFC 9110: HTTP Semantics — idempotency as a property of an operation rather than of a client's intent, and the basis for an effect-ledger key.