It is quite uncontroversial to claim that different mixes can sound different. (Which isn't the same as to say that when you think you hear differences there are differences, because even when you are convinced the mixes are different, you can surely be fooled the same way as when you know the formats are different.)
Also, I am so damn sure that 192/24 is not enough ... yesterday I produced a proof of concept without even attempting it.
You wanna see the ABX log you say?
Did you notice I wrote "192" and not "192000"? Neither did I, when I did it! Backstory: Since TAK doesn't handle 384 kHz, I tried to make an 192 kHz file to see how TAK would compress it compared to others. Only to find that TAK still refused. Only because it doesn't handle 192 Hz either. (Other codecs do ...!)
My point being: clumsy people like myself should only do lossy compression/conversion for the purpose of creating disposable files. Not for throwing away the source.
If some hi-rez file sounds better because it is mixed different, then I'd say, don't downmix it to save space, unless you are willing to take the risk you might destroy something. Which is likely even without typos, if you don't know precisely what various software might do. 32 bits is overkill you say? Still do not ffmpeg -i 32bitfile.wav outfile.wav, it will clip if the 32bitfile is float exceeding 0dB.