Gregory, welcome to the forum!
I definitely cover this topic in the third edition of my book, which I sent you just the other day :-). And soon, my talk I just gave to the San Francisco AES section will be available on YouTube.
While you’re waiting till the video become available, I’ll tell you that in my studio, I routinely upsample any single sample rate mixes to double sample rate at 32 bit float before beginning mastering. Keeping in mind that your DAW will calculate at its internal wordlength/precision regardless of what is input, I’m comfortable saying that if you upsample to 96k at 32 bit float, it should be fine. If your DAW can accept 64 bit float and your SRC can output at 64 bit float I see no harm, maybe some wasted space, but it shouldn’t be harmful, though I doubt the difference will be audible. Certainly not intuitively obvious to the most casual observer. 🙂
Reaper reportedly processes at 64 bit float, so for example, if you import a 32 bit float file Reaper will output 64 bit float for any multiply.
Hope this helps,