Just an idea. When running workflows on a Virtual machine, sometimes the VM lags and need to wait until it can continue. This is often the root case of errors, e.g. writing and closing an Excel sheet several times can work fine, but suddenly it needs 10 sec just to save it. Then it crashes at the next write.
So is it possible to have a I/O or CPU check that triggers a delay (pause the workflow) when the VM is too active?