I suppose it could be an interesting experiment.
If the streaming is lossless, it should be equivalent to just playing the file [unless the streaming site imposes a limiter on maximum peak levels).
For lossy encoders, there is a plugin out there that lets you hear various lossy encoders (MP3, OGG, etc.) right in your DAW.
Since perceived loudness is only indirectly related to peak level (as transformed by crest factor and a few other considerations), it makes sense to manage peak and LUFS levels to fit in the available dynamic range of the medium and the dynamics of the music itself.
Personally I keep peaks on master mixes at around -1.5 dBFS TP, which is conservative, but has never caused any clipping with a lossy encoder I’ve tried. A separate concern is listener perceived loudness, which is most accurately measured by LUFS values.
I typically have masters between -13 and -15 LUFS-I, under the theory that it’s better to listener turn you up because they want to hear you, rather than turn you down because they find you annoying or overpowering.