arcpy crashing mid-process

[JamieKass]

I've had multiple experiences where arcpy crashes for no apparent reason as I'm running my code. Usually, it happens when I attempt to run a tool on the product of another tool. This time around, the problem is running MultipartToSinglepart on a Union product in a feature dataset, attempting to output to a different file geodatabase. This step occurs after I've completed the entire process for one feature class, and I am attempting to run it again on another one with different names. When I run this particular case again after the Python engine has been restarted, it runs to completion. As the Python engine itself crashes when it errors, I get no informative message as to what occurred and why.
For some background on my code, as I have taken a few safety measures already, I delete all cursor objects, delete all layers after their utility has run up, and am working in a file geodatabase. Is there anything else that may be lingering in my file geodatabase after some processing that will make Python crash so hard? If there is a fix, I'd like to try and implement it.

[MichaelVolz]

Can you add some debugging to your python script to see if variables are getting passed the expected values, especially after the initial loop of your script?

[JamieKass]

I managed to get around this particular problem by throwing in a couple of RepairGeometry's after some geoprocessing steps. This technique works I suppose, but it brings up a new question: why does ArcGIS produce geometry errors after operations like Erase and Union? Shouldn't the product be free of errors? I suppose there is no internal error checking before the output is returned.