Too friggin drunk to read all the hard core advice here, so sorry if this is already covered. My opinion lay below layman, so may not be worth a hill of compost in the bigger, more informed picture.
Maybe depends on what the target of the end product is.
If you are translating to another system with thousand dollar monitors that need every nuance of a 192/24 or floating 32bit representation, then you got to go with the big guns.
Some claim that 24 bit depth can record trails of faint reverbs or the like that merit the argument to record at higher bit rates.. Reference the depth of the noise floor. Twice, three or way more better than what the ancient rock and roll masters had struggling with an Ampex tape machine. 56 db s/n ratio was all they had, but they did magic.
If the target is a CD boombox, you are stuck with the 44.1k/16 bit protocol. Everything you have oversampled is lost. Noise floor comes up to 16 bit obscuring the silent subtleties. All connect the dots points on the digital timeline are thrown out down sampling to 44.1 k.
Anything oversampled to get pristine quality has to be degraded to make the CD. Anything undersampled has to be bloated up to an empty space with no added data.
This why in my ignorance, since my target is eventually a CD, record at 44.1k and 16 bit depth, just do it the way it ultimately has to be restructured to make the CD and not having to worry about artifacts in the conversion.
Something I found. My old PCI Delta 1010-LT allows recording at 22.5 k or whatever close. Gets beyond a curse of pristine digital unforgiveness and seems to throw in a bit of old analog goodness.
My opinion, if you are going to publish a CD, do 44.1 k and 16 bit depth. All else upper and lower is lost until they update the CD standard, and you are at the mercy of the integrity of whatever you have chosen to up sample or downsample your final mix.
My two cents.
John