I record at 24bit, 48kHz. But not because I can of any quality difference between 48 and 96, but because and the files are so much smaller at 48 kHz. They copy to the computer fast. They render much faster.
I record at 24 bit for added headroom, But not because of any musical content with over 16bits of dynamic range, but becauseI can run levels a little lower for safety, and know that I'm not going to clip- and that I'll normalize in post.
I release two versions, one at 24bit / 48kHz and one at 16bit / 44.1kHz.
The former is a stereo-balanced and normalized version of the raw data, for Audiophiles, purists, and people doing MTX with released SBD versions. The latter is a Redbook compatible, SBE fixed source. This gets EQ, Dynamics work, Stereo processing, and some **very** subtle "analog" saturation like one might get from transformers or tubes. It sounds better to most people without fancy stereos or headphones. That's also the version that goes on my phone for the subway or the car.
This strikes me as the best of both worlds, an "audiophile," hi-fi version, to make me feel good about myself and my gear; and a version which should sound really good, even to people who grew up in loudness wartime, and who are listening over earbuds or computer speakers.
Could I pick out the difference between the 24/48 and a straight 16/44.1 conversion in a ABX test? I doubt it.