On-Set Performance vs. Audio Quality: When to punt to ADR?

Hey everyone,

I'm wrestling with a common dilemma on a short film I'm editing. We had a really powerful, unrepeatable on-set performance, shot on an ALEXA Mini with a couple of SkyPanel S60-Cs lighting the scene. Unfortunately, the location was noisier than anticipated, and despite our boom op doing their best with an MKH 8060 into a MixPre-10 II, the dialogue has some significant background noise.

I've been experimenting with some AI-powered audio repair tools, and they're definitely improving things, but there’s still a noticeable artifacting on certain words. My director is thrilled with the raw performance, the actor really nailed it. The alternative is ADR, which could clean it up perfectly, but I'm worried about losing that raw, in-the-moment authenticity.

The 'guides' always present salvaging audio and recording ADR as two distinct options. In a real-world scenario where you have a noisy but genuinely powerful on-set performance, what's your ultimate metric for making that final call to scrap the field audio and go for ADR? Is it a specific percentage of intelligibility loss? An emotional impact threshold? Any advice is much appreciated!