I've noticed newer proof engineers quickly get distracted by one of two things:
1. trying to set up the perfect definitions, lemmas, and theorems from the start, or
2. trying to game the proof of the top-level theorem that's already there without thinking much about what is needed to get there.
It occurs to me that both of these are problems in the balance of bottom-up versus top-down proving. Experienced proof engineers are more likely to smoothly move between these when it will give them the most information to make progress.