It's great for sound recording and mixing, since it has so much headroom and dynamic range. Bits are cheap enough, why not even go to 32-bit floats? With just 24 bits, you can even do silly things like digital volume control, and then get rid of your preamp.
BTW, for consumer playback, it's not 24-bits so much as the 96 kHz that's key. And the main reason was that you could use cheaper anti-alias filters with a really wide transition band.
Some labels like Nuclear Blast offer 24/44 albums but not 24/96. I still hear the difference in 24/44. It's sometimes only $4 more for the 24-bit and I have the storage, and I do this for my favorite bands like Avantasia. I'll pay the $4 to help me get the better version regardless of the 16-bit/24-bit, especially if there's a chance that 24-bit has a better master.
32-bit actually might be a good idea, but I doubt it would be noticeable at all. Basically the number range for 16-bit is -32,768 to 32,767. Going up to 24-bit brings that up to -8 million to 8 million (256 times the resolution). 32-bit is -2 billion to 2 billion if integer, even more insane for float. It goes from a ridiculous amount of discreet values to an even more ridiculous amount. For me, it's not about the range, but about the resolution.
Here's what I say about 16-bit to 24-bit, especially with that kind of difference in discreet values. If you say in terms of technology or software that you will never, ever hear a difference, my BS detector goes off. I code software, and there's too much weird stuff that goes on with software and hardware for me to believe when people use absolutes.
We have processors that handle 32-bit easier than 24-bit. Microsoft's game audio libraries actually operate in 32-bit float if you start doing complicated things with audio.