Every trigger writes a result file into processed/ and nothing in the repo deletes them. One machine held 83 results, 75 of them written before the submitted field existed.
This is recorded as a decision rather than an oversight — it was noticed while fixing the trigger watcher and deliberately left. Filing it so the decision is visible instead of living in a commit body.
Whatever policy is chosen has to keep a result readable for long enough that the caller can still fetch it: the harness reads <uuid>.result.json by name, and a chain can take minutes.
Every trigger writes a result file into
processed/and nothing in the repo deletes them. One machine held 83 results, 75 of them written before thesubmittedfield existed.This is recorded as a decision rather than an oversight — it was noticed while fixing the trigger watcher and deliberately left. Filing it so the decision is visible instead of living in a commit body.
Whatever policy is chosen has to keep a result readable for long enough that the caller can still fetch it: the harness reads
<uuid>.result.jsonby name, and a chain can take minutes.