Skip to content

test(e2e): make bounded finalization deferral scenario reproducible #696

Description

@moncher-dev

Problem

bash e2e/run-e2e.sh api-progress-unknown 90 consistently records one run-finalization-deferred event, but the runner asserts the bounded sequence contains three events.

The runtime only enters processRunExit finalization when the worker exits. After the first missing-item deferral, HTTP POST /api/v1/refresh requests reconcile the project but do not invoke worker-exit finalization again, so the counter cannot reach its bound in the current black-box scenario.

Reproduction

bash e2e/run-e2e.sh api-progress-unknown 90

Observed event: consecutiveDeferrals: 1, maxDeferrals: 3, exhausted: false.

Expected

Either drive the runtime through three genuine finalization attempts in the Docker scenario, or revise the scenario so its assertion matches the supported lifecycle while retaining unit coverage for the bounded retry sequence.

Discovered while revalidating #692; Docker isolation itself is verified by concurrent worktree runs.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions