[ScriptedAlchemy]

Summary

Step 5 of the storage redesign (docs/plans/storage/README.md), second slice. Pause and resume of a Work run now run as one writer job.

Before this change, a pause made three separate trips to storage: the run's frontier on a reader, the workflow journal replay through a begin_deferred writer transaction, and the compare-and-swap publication in its own interactive transaction. Every trip parked the request thread, which is why Work requests ran under off_native_worker. Now RegisteredWorkApplicationServicesV1::pause_run and resume_run run the unchanged WorkRunControlService decision inside WorkSqliteStorage::transact. The frontier read, the journal read, the decision and the publication commit together or not at all, and the awaiting task parks no thread. offer_work_blocked_interval_receipts still runs after the commit.

This is the shape the rest of step 5 follows. The decision stays a synchronous contracts function, and it runs on the writer thread over a storage scoped to that job's transaction.

• WorkRunControlTransitionPort (contracts) holds the three methods a transition uses. It has no Send + Sync bound, so WorkSqliteJobStorage, which borrows the job transaction, can implement it. The rest of WorkRunControlStoragePort is unchanged.
• WorkSqliteJobStorage holds no ExactSqlHandle. A job therefore cannot call back into the writer thread that runs it.
• The handle path and the job path share every SQL body, through RegisteredWorkQuery (now also implemented for ExactSqlJobTransaction) and a new RegisteredWorkWrite.
• The handle-based transition stays for its one remaining caller, synchronize_fan_out_run_controls (synchronous workflow fan-out). It goes when that path migrates.

Test plan

• work_run_control_jobs::pause_and_resume_awaited_on_a_native_worker_park_no_native_thread (application_suite): pause then resume on a native worker leave the native writer-wait and reader-wait counters unchanged. Putting back the old synchronous call in pause_run fails it (writer waits left 11), and so does the same change in resume_run.
• work_run_control_storage::a_run_control_job_publishes_with_its_frontier_read_or_not_at_all (rusqlite_suite): a job that publishes and then refuses leaves no control row, and an accepted job commits it. Committing regardless of the verdict fails it.
• bazel test --config=clippy for rusqlite-runtime unit_test and rusqlite_suite, contracts unit_test and contracts_suite, application unit_test and application_suite, and daemon-service unit_test
• rustfmt_check_2024 for each touched crate
• //crates/tracedecay:daemon_suite, :mcp_suite, :session_suite, :unit_test in CI

Refs #4052

---

[changeset-bot]

⚠️ No Changeset found

Latest commit: 71f9667

Merging this PR will not cause a version bump for any packages. If these changes should not result in a new version, you're good to go. If these changes should result in a version bump, you need to add a changeset.

Click here to learn what changesets are, and how to add one.

Click here if you're a maintainer who wants to add a changeset to this PR

[devin-ai-integration]

🔍 Devin Review: 1 flag

Not posted on this PR by your GitHub settings — view it in Devin Review. (Configure)