So, basically, the windows/OS PC volume, keep just shy of 100%?
well that's my first concern when nothing else seems to be an issue, but it deals only with some potential(and usually small) clipping. that's why I think the quest for fidelity requires some measurements so that we can, when necessary treat the issues we find.
if I take my situation, changing the volume level on my O2 doesn't change the signal quality much. and the noise floor is still not audible in the headphone with the digital level maxed out and the amp set to my preferred loudness . but I can get up to a good 0.5dB of channel imbalance even at the lowest gain. to reduce that I could lower the digital volume level until I'm out of the first 1/3 of the amp's knob where the problem is most significant, or like I did I can fool around with the knob and a voltmeter until I hopefully get lucky and find a small channel imbalance within the area where the loudness would be fine for me even with the computer at 100%. so now I leave my amp alone where I get a little less than 0.1dB imbalance, and I set the loudness only from the computer. it's not ideal, I just had 2 variables and decided channel imbalance was my main issue at this time.
on my portable amp, I have a digital control of the volume(analog volume, controlled by a chipset, not really digital) and thanks to that, channel imbalance is virtually non existent at all volume settings. so here I set the loudness with the amp's knob and do keep my digital level short of 100%(for intersample clipping paranoia). it's also beneficial that way because on that specific amp, the noise increases as I increase the amp's volume, so here I have nothing to "gain"(pun forbidden by the Geneva convention) lowering the digital level and pushing the analog one. also I would soon hit a wall with the hd650 as the amp reaches 1%THD a little over 50mW into 300ohm. so while I have enough for any loudness level I can dream of using, I can't just give digital level away like it will never matter.
we could on rare occasions fall onto other situations. maybe the DAC outputs 2 or 3V but you're using a portable amp that was designed to get about 1V(happened a lot when the ipod was all the rage), then you may need to decrease the digital level a little just to start getting a listenable sound.
and of course one might use properly mastered albums where the guy didn't stick the signal to 0dB like a noob. or maybe you're using replay gain with careful settings(the slow scanning ones with a good oversampling rate and avoid clipping option) so all the brickwalled stuff end up digitally attenuated anyway. in which case you might not have any file clipping and there is no point avoiding 100% digital volume level anymore. the end result is the same, someone or something placed the signal so that it's not too close to 0dB.
different devices and situations give different problems(or no problem hopefully).
I your oversampling filter uses more bits than the signal e.g. 32 bits for 16/24 bit material (who has 32 bit music - seriously ...) then intersample clipping should not occur. (?)
So go source 100%.
I don't see why that would change anything for intersample clipping? bit depth does nothing for the 0dB upper limit and crappy sound engineers.
am I getting off topic enough? ^_^ soz.