Kojira does a number of things. It can sometimes get bogged down crunching data and take too long to come back around to the part where it actually fires off newRepo tasks. This can keep kojira from maintaining the configured maximum number of jobs when there is a backlog.

This problem might manifest when there are a large number build tags defined, and can be aggravated by:

1. a very large number of repos on disk
2. slow reads on /mnt/koji
3. slow queries on the hub