[Hua Peng (StataCorp)]

That's what we suspected. I did some tests with mocked up cases using SQL database (we sent a request for a trial license but are still waiting to hear back from esri) and can not reproduce the issue.

[Mike LeGower]

I'll let you know how broader testing beyond toy datasets goes. I suspect it was mostly Esri issues that caused unexpected Python exceptions with the added irritation of bad garbage collection after those events that caused file locks to persist and prevent overwrites on subsequent runs. Eliminating the arcpy exceptions seems to have mitigated the garbage collection issues.