Model Builder Version Problems?

[ChrisMcEndree]

Hey folks! My team is experiencing a strange problem where the same tool used on different versions produces different results. We are running several select by queries in a row to help with the quality control process, and when some of the team members on 3.7 use it, it selects far too many/the wrong results. When users on 3.54 use it, the tools seem to work as intended and produces expected results. Not sure if I'm allowed to share the actual model for proprietary reasons, but I wanted to know if anyone else was getting different results version to version. 

[MErikReedAugusta]

I know you said you can't share the actual model, but can you share at least the GP Tools you're calling from inside it?

I was just helping a user on another platform with a similar case of different versions producing different results.  We weren't able to get enough of their code (also because of proprietary issues) to nail down the specific problem, but after a ton of back and forth, the general problem seemed to be that some tool or library that their code relied on was changed between the versions, and started producing data they hadn't accounted for.  In their case, they were suddenly ending up with divide-by-zero errors, because they suddenly started getting a zero value from some chunk of code that had only ever returned valid, nonzero values before the update.

If one of the tools you're using was updated during that last update pass, you might be passing around data that's different from what you expect, which is ultimately leading to the downstream change you're seeing in the results.