It has become too common recently, even when the models used are same.

I have multiple examples. Just yesterday, when I was trying to implement a caching system for my website Nasengetu, and it completely stuck in a loop-where it runs a log fetch using the git command and then thinks about the result, then runs the command again, and think again. And this repeated until I stopped it when it almost completely exhausted my five hour request limit.