Find the affected run
Start with timestamp, input record, and outcome. Read the run status before assuming every unsuccessful run is the same kind of failure.
Errored, safely halted, held, scheduled, and handled runs mean different things. Diagnose the run state, then verify whether the intended customer or operational outcome actually happened.
Start with timestamp, input record, and outcome. Read the run status before assuming every unsuccessful run is the same kind of failure.
Compare step input, output, error message, and—when available—the HTTP method, endpoint, response, and status code.
Connection loss, task limits, flood protection, rate limits, app outages, and changed required fields require different responses.
Verify whether earlier steps already sent, created, or updated something before replaying the run.
A successful replay resolves one execution. Dependability requires knowing why it failed, which records were affected, and how the next failure will be surfaced.
A bounded repair may involve mapping validation, connection repair, explicit stop conditions, retry limits, owner alerts, or a safe replay procedure. It should not silently broaden into an open-ended rebuild.
Augmentive AI first confirms fit, then separately scopes a paid diagnostic. Repair is quoted only after the evidence supports it.
Describe the run status and business consequence. Do not send credentials through the form.
Request fit review →