32-bit Float & AI NR: Has your on-set gain staging changed?

Hey everyone, I'm a Casting Associate, and I've been really curious about something regarding on-set audio lately. We just wrapped a commercial where our sound mixer was using a lot of 32-bit float recording capabilities, and he mentioned how much it changed his workflow. I'm also seeing a lot more about AI-powered noise reduction like iZotope RX being used in post.

Historically, 'clean' audio meant meticulous gain staging to avoid clipping or getting too close to the noise floor. We’re often shooting on an AMIRA or a VENICE 2, sometimes an AMIRA, with MKH 416s or MKH 50s, so good practices are usually ingrained. But now, with 32-bit float offering such a massive dynamic range, and RX seemingly able to salvage so much, has your fundamental on-set approach to gain staging and capturing 'clean' audio changed at all? Are you still as strict, or are you taking more risks knowing what can be done in post?