You record at higher sample rate because of math when mixing two waveform representations i.e.:
"What is the result of 1.4 * 1.3 ?" compared to "What is the result of 1 * 1 ?"
Well in this situation, 1 * 1 equals 1 or 2, the uncertainty of which can become an audible artifact when iteratively repeated (as might happen in some recording pipeline) but on a scale of 1-to-22'000+ it matters not at all if it was 1 or 2, just that it was some kind of value in that range. No percentage of human ears can discern between 48kHz and higher sample rates, and a statistically minor percentage of human ears can reliably discern between 44.1kHz and 48kHz. Sadly there is this large catalog of audio CD media which was mastered in this almost-but-not-quite ideal 44.1kHz sample rate so we're here decades later arguing about it yet.
As for bit-depth resolution 24-bit is ridiculous but 16-bit is getting pushed to the limit by the loudness war and so the idea that a carefully mastered recording will happily sit in 16-bit sample size is not justifiable. What you'd want is 20-bit however everything about that is an enormous pain in the neck to implement, so the extra odd 46kb per second is acceptable to get things that divide evenly by 16. You cannot hear this difference if the audio mastering is done properly, but by god you will hear everything wrong with it because morons have taken over. With 24-bit (or 32-bit) depth there are ample ways for an audio mastering process to fail but with 16-bit depth it's pretty much guaranteed that it will sound like garbage unless the audio mastering is done by a genius with total creative control.