The issue is related to the process dependencies resolution. It is executed every time the engine is paused and then resumed. The big number of processes (mote than 150 in my case) is causing trasaction timeout, because dependencies resolution is executed for all the process in one transaction. The default transaction time limit is set to 60 seconds and can be increased. However it can be increased only globally and that can be bad for stability and performance. Do you think it is reasonable to break this single transaction to separate transactions for every process and resuming tenat as well?
You can workaround this problem by setting timeout to higher value. After resuming tenant it can be decreased. However it requires server restarts.