If the poller master doesn't complete in time and ends up with overlapping jobs, neither job will ever complete because both running jobs will think they are the master and wait for the remaining node to exit.
Do not allow the master to join twice
Auto-Deploy finished, Test PR at http://1377.ci.librenms.org or https://1377.ci.librenms.org
deployed to a two poller node setup with 2600+ devices. Will feedback in a day or two.
Ok, all running ok and I don't see any issues with the code.