32-bit float & AI NR: Has on-set gain staging changed for you?

Hey everyone, David here. I'm cutting a short right now shot on an AMIRA and I'm really trying to maximize our audio quality. With tools like 32-bit float recorders (I've got a MixPre-10 II) and the incredible advances in AI noise reduction like iZotope RX, I'm genuinely curious: has your on-set approach to gain staging fundamentally shifted, or are you operating pretty much the same as before?

I've been experimenting with backing off the faders more, trusting that 32-bit float will catch everything, and then relying on RX in post if something unexpected crops up. But I'm wondering if I'm missing something, or if the 'old ways' of meticulous gain setting are still paramount no matter the tech. What's your current philosophy here?