Coin gets stuck after running out of python threads

After couple of hours of every Coin restart, the service stops responding and starting up integrations, which means the CI is completely blocked.

Based on analysis, there is a reasonable suspicion that the reason is too many threads left behind by the scheduler process. Counting threads for scheduler process results in

vmbuilder@vmbuilder:~/qt-ci$ ps -T -p 25179 | wc -l
4967

Eventually this causes python process to refuse creating new threads because it does not have enough resources to do so.