ImprovedAPI
Deployments that crashed, ran out of memory, or couldn't pull their image now count as failed. GET /v2/deployments?status=failed returns them and counts them under failed, GET /v2/services shows their service as failed, and GET /v2/projects counts that service in failing_service_count. Before, they matched no status: a status=failed filter left them out, and their service could read as running.