Durable Lifecycle Execution

Context

These operations outlast a request and need inspectable progress and recovery. Execution must not begin against records that might roll back.

Decision

Use shared durable jobs for construction, deletion watching, purge and reset, committing each new job before starting its execution.

Alternatives Considered

Run the work inside the request: Request-bound execution can time out after some work has committed. Start the execution inside the transaction: External execution cannot participate in database rollback. Queue the work: The work needs ordered steps, waits and recorded progress beyond a single message.

Consequences

Normal submissions store their job with domain changes; provisioning retry retains a separate-commit gap. Job failure behavior is declared for workflow and reconciliation use. Shared job inspection supplies progress, events and retry lineage.