kandev/docs/plans/step-entry-dispatch-convergence/task-08-coalescing-exclusion.md

id, title, status, wave, depends_on, plan, requirements, acceptance_criteria, system_design

id	title	status	wave	depends_on	plan	requirements	acceptance_criteria	system_design
08-coalescing-exclusion	Entry-triggered runs bypass coalescing	pending	8	03-ledger-ownership	plan.md	REQ-OFFICE-STEP-ENTRY-DISPATCH-005	AC-OFFICE-STEP-ENTRY-DISPATCH-005.1 AC-OFFICE-STEP-ENTRY-DISPATCH-005.2 AC-OFFICE-STEP-ENTRY-DISPATCH-005.3 AC-OFFICE-STEP-ENTRY-DISPATCH-005.4 AC-OFFICE-STEP-ENTRY-DISPATCH-005.5 AC-OFFICE-STEP-ENTRY-DISPATCH-005.6	../../specs/office/system-design/step-entry-dispatch-convergence.md

Task 08: Entry-triggered runs bypass coalescing

Summary

Mark runs enqueued by a step-entry action so they neither merge into a queued run nor are chosen as a merge target. The flag is carried both in memory, for shouldCoalesceRun, and on the run row, because CoalesceRun's subquery reads persisted columns only.

In scope

• EntryTriggered bool on QueueRunRequest in both internal/runs/service and internal/workflow/engine, which must stay identical.
• The engine setting it for every run enqueued by a step-entry action, and shouldCoalesceRun returning false when it is set.
• runsServiceEngineAdapter.QueueRun (internal/backendapp/main.go) copying EntryTriggered from the engine's QueueRunRequest onto the runs-service request — it currently copies every other field, so an unpropagated flag here would silently discard the engine's exclusion before persistence ever sees it.
• The r.migrate.Apply("runs.entry_triggered", ...) ALTER in the office repository, the insert path writing it, and CoalesceRun's subquery gaining AND entry_triggered = 0.
• The columnExists probe, falling back to pre-requirement coalescing with one ERROR when the column is absent.

Out of scope

• Any change to the idempotency key or to the task-comment prefix. Idempotency suppression is unaffected.

Acceptance

• Two entry-triggered runs for one (agent_profile_id, reason) inside the window both remain runnable.
• A queued entry-triggered row is not selected by CoalesceRun as the target of a later ordinary request.
• With the column absent, run enqueue still succeeds and the exclusion is inactive.

Verification

cd apps/backend && go test ./internal/runs/service/... ./internal/office/repository/sqlite/... ./internal/workflow/engine/...

Files likely touched

• apps/backend/internal/runs/service/service.go
• apps/backend/internal/runs/service/service_test.go
• apps/backend/internal/workflow/engine/adapters.go
• apps/backend/internal/office/repository/sqlite/base.go
• apps/backend/internal/office/repository/sqlite/runs_test.go
• apps/backend/internal/backendapp/main.go (runsServiceEngineAdapter.QueueRun must copy EntryTriggered through)

Dependencies

Task 03, which establishes which dispatcher enqueues an entry-triggered run.

Risks

• Asserting only that an entry run does not coalesce passes against a service-side-only implementation, which is the half that is easy to ship alone. The target direction must be covered.
• SQL naming a missing column fails the whole statement, so an unprobed entry_triggered breaks every run enqueue rather than only the exclusion.

Parallelism

sequential

Inputs

• System design, "Coalescing exclusion" and "Migrated columns are probed before use".
• runs.outcome in internal/office/repository/sqlite/run_outcome_activation.go as the precedent for both the ALTER site and the probe shape.

Results

Pending.