dan-da
commented
8 months ago I think that pausing mining would be a band-aid for the real problem, whatever it is, which should be fixed instead.
It is possible that the problem is already fixed in more recent code. In particular, the locking has been re-worked and is much simpler now.
Leaving open for now. Will attempt to reproduce another time.
When mining (without throttling), one CPU is used at full capacity. When initiating a transaction, a proof is generated and all CPUs are needed. So it may cause scheduling conflicts when we are mining (without throttling) and initiating a transaction. I have seen a node crash when attempting to do this and I suspect scheduling conflicts is the problem.