In a less dramatic view, it can be a good idea to have the amp set relatively near your usual listening level(a little above that) so you don’t end up with 50dB of digital attenuation that might not be as audibly transparent as we’d like. The obvious drawback being that anytime you want to crank up the music, you may have to get up and adjust the amp. If you’re using some options like replay gain or similar that makes most tracks subjectively feel as loud as others, then my proposition can be pretty convenient. Just keep in mind that such a loudness matching option is adding some more digital attenuation to tracks. Some choices have to be made.
I know very little about the streaming solutions you mentioned, but as a rule of thumb, if you use digital attenuation, it’s not a bad idea to have the player send 24bit or more to the DAC when possible. That has nothing to do with the music file itself being 24bit, it’s only to have more digital room to reduce the amount of truncation. At 16 bit any attenuation will discard the lower bits of the signal. Usually that’s noise and the very end of some room reverb so it’s ok IMO, but the more attenuation, the more you might end up cutting off a relevant quiet part of the music. The vast majority of music will for a reason or another have 60dB or less of real dynamic(mixing, compression, or just the band not playing crazy loud in a room with some unavoidable noises). So in theory and for me at least in practice, you can get away with 20 or 25dB of attenuation most of the time. Might not stay true for very dynamic classical played loud on very quiet passages that last a while.
Personally I have digital attenuation on top of digital attenuation on top of... and I’m still a happy camper. I always have replay gain, some EQ, some heavy processing in a speaker simulator, and I adjust the playback level in Foobar2000(1700% digital). TBH when testing in 16bit, to my own surprise I still often couldn’t tell as I don’t listen to music loudly and my amp is set near my usual listening level. I have my output set to 24 bit anyway because I know it’s a more accurate solution and because depending on the EQ(to go with a given headphone or IEM), I sometimes end with tracks lowered by 25dB before I even start adjusting the listening level digitally. It just wouldn’t be reasonable to stick to 16bit reaching my DAC under those extreme conditions, but otherwise, I’m the type to feel like comfort and convenience can easily make up for fidelity when measuring my enjoyment of music
