""Hey that's a weird thing in the result that hints at some other vector for this thing we should look at." "

This is very common already in AI.

Just look at the internal reasoning of any high thinking model, the trace is full of those chains of thought.