Luis tells me that proposals set to auto-calibrate migrated runs have a cron job which will launch calibration for any runs with calibration status 'NA' (not available). We've been setting this when calibration fails, which could lead to an infinite loop of calibration failing and restarting. I don't think we've seen this yet - we fixed a different infinite loop in !796 (merged) - but now that we have a separate 'E' for error status, we should use that instead.
Update: this now also uses the new 'IP' (In Progress) status for running jobs, instead of 'R'. The myMdC change was deployed on Saturday (2023-02-11).
Deployed on max-exfl017, see below.