Hey folks, this is
@WillVT 's brother that he mentioned. I run a recording studio in Austin. So yes, the main "aggregator's" as they are called only accept 16bit-44.1khz wav files. Some of the main players are Tunecore, CD Baby, and Distrokid. I have used all three. Last I checked (which was not long ago), they are only accepting the aforementioned bit depth and sample rate. I can confirm that from a release I personally uploaded on CD Baby just over a month ago. On the interface these services provide, you choose which streaming services are delivered to, which is a lot more than just the big players like Spotify, Apple, Amazon, Youtube. We never have any direct contact with those services-- you just say when you want it to drop, and it appears on those services on that day. Good luck ever getting a human that would have a clue what you're talking about from tech giants like these.... Major label artists have a different path which does allow 24bit, but I wouldn't be surprised at all if many deliver 16-44. Many labels and artists do not understand or care.
Just to provide more info from this side of things, I am mainly a tracking and mixing engineer, but I also do mastering. I probably master 10 records a year. For stuff that I mix and personal projects (like this one that started this discussion with my brother), I usually push to send it to a dedicated mastering engineer. I recently been using JJ Golden of Golden Mastering in LA. Like most mastering houses, the standard product you receive is 16-44.1. You can pay extra for an MFIT (mastered for itunes) which is 24b-48k. As far as I know, Apple is the only one that has an actual standard for mastering engineers to adhere to, although even that is kinda BS. I also deliver 16-44.1 to clients when I master. It's pretty standard to deliver 24bit and whatever sample rate the project was recorded at for vinyl.
Now the controversial part. In the engineer world, it's pretty well understood that higher sample rates are only of benefit for processing in the digital realm, aka plugins. Plugins that don't oversample internally can kick back harmonics from the hard cut on the high frequency (at 22khz at 44.1 sample rate). So it's best practice to use higher sample rates for processing, but of questionable benefit for listening. I know of at least one mastering engineer that will upsample from 44 or 48 to 96khz for the mastering processing. When I say of questionable benefit for listening, it's because most humans can't hear past 16khz, and pretty much none past 20khz....not to mention that it degrades with age. So 44.1 hitting 22khz is already beyond what we can hear. Personally, I record at 48khz or 96khz-- but I get a lot of stuff in from other folks at 44. I don't know anybody that records at 192. All that said, my tape machines go above 22khz, and drop off gently. I do have a theory that those frequencies above the audible range do have an effect, but then you get into are your speakers even reproducing it-- if they have DSP, it's only to the sample rate they are using.
edit: Also, from a quick google it seems many of the players for Amazon HD do not report sample rate correctly