Now they're getting paid, so fuck you!

That exactly tracks. You can't feed answers from an AI into an AI. It gets all incesty (technical term). So they have to ban user submitted AI answers.

Were they trying to avoid having AI produced output sold as LLM input along with their human user generated content? I wonder if this was some big picture decision or pure coincidence.

Banning AI answers was reasonable though. People were posting were too many not verified and incorrect code snippets that entire quality of the platform would decrease. AI still makes more mistakes than humans that provide responses on stack.