I still on the fence of which sample rate and bit depths are the ultimate choice, but let me just put my understanding, ideas and observation here.
For recording, I might record at 96/32 or 96/24. I've never tried it though. Recording guitar or realistical sound from microphone would have better accuracy with higher sampling rate, so I might pick 96khz.
For any General setting, especially for Cakewalk, I'd set 32bit for bit depth. The reason is, the inside a DAW, every process adds extra bits. And some vst use either 32 float or 64 double precision. So the internal processes are all done with 32 or 64 bit information. If you set 24 bit depth, when you freeze it or record it, some information will be trancated. So even though internal process are 32 float or 64 bit double precision, once you write a wave information then it will be tranced. So the sound changes with bouncing it or when you freeze it. Processing it multiple time may deteriorate the sound by occasionally lower the bit depth to 24 bit during the work, which I think I can avoid by setting it 32bit for DAW's file recording. (For the program itself, I use 64 double precision engine, even though some says 64bit double float doesn't makes the sound better as 32 float is more than enough)
People may not notice the difference but in theory, transaction happens.
For mastering, without a doubt, 96/32 or higher, maybe 192 if it's an option. For mastering, for better clearity with less distortion, higher sampling rate is generally introduced by mastering engineers too. Bob Katz says upsampling to higher sampling rate for mastering improves the quality. I'd skip the theory of why upsampling is good for mastering but I think it comes with some downside. Eventually, the file needs downsampling to 44khz, you have to use filter to remove folding noise, and that's how resampling works. But the problem is, whenever you use a filter, the peak changes. So for example, before resampling it, the peak was set to -0.3db with blickwall limiter but once you downsampled to 44hz, then it could exceed the peak of even 0db! so this may cause other distortion so I need to take an extra care about the peak with SRC. Reintroducing a limiter at downsampling process may change the sound quality, too.
For production, I think it's difficult to say which sampling rate is the best. Most of the sample library offers the samples with 44hz. So with different sampling rate setting for a DAW cause resampling of these sample files inside the DAW. The inside DAW resampling doesn't come with the good quality compared to iZotope SRC.
And for synth, the sound changes depending on sampling rate. Most of the time, (not always, thought) higher sampling rate would make synth sound better. So it may be good to use higher frequency settings when the synth doesn't offer oversampling options inside.
When you use distortion, like tube distortion inside z3ta, the sound clearity differs by the sampling rate.
And sometimes, higher sampling rate produces more Inter Modulation distortion, while 44khz produce more potential folding distortion. Some times for some reasons, idk why but folding distortion makes the sound warm. And for IMD, by introducing linear phase filters that cut higher frequency above in audible frequency helps make the sound even clearer, but the filter might potentially cause pre ringing effect, I don't think it's audible nor significant but I think it might change some transient information, or affect lower frequency as well. Which in the end will be added up to the entire mix to some extent. So it might have some weird time shift in sound.
So as the sample comes with 44hz, producing at 44hz might be a fine choice. But I don't know.
Maybe 48hz is better as you also use many effects.
And for mixing, you probably need more plugins to be introduced, 48hz might be the best as you can reduce both IMD and folding noise.
If you set 88hz or higher, the sound is so clean with less distortion but now I miss some lowend energy, making me think weather this was a better decision or not. Folding noise is not musical in theory but the extra clearity doesn't mean the song is warm at low frequency, nor powerful.
And plugins work differently too. Many EQs will appreciate with higher sampling rate. and also non liner processes like compressors or distortion types of plugins can sound differently.
For clearity, higher sampling rate would be better but some plugin sounds better at 44khz or 48khz. It'd make sense, though, that the developer forcuses on rather 44hz or 48khz as that's a common sampling rate. Or, maybe my ears are not good. Sometimes they even inroduces oversampling options inside a plugin so 44 or 48 are not a bad choice.
So maybe, maybe I will use 48khz for mixing as it's good on both CPU and the quality.
I think for recording from mic, maybe 96khz or 48khz, 48khz can be realistic for both production and mixing. And 96khz or higher for mastering.
Well but I don't know...