Well, look at it this way - once you've got your tone converted to a digital signal, volume becomes irrelevant. Turning it up or turning it down becomes a mathematical equation rather than a gain stage, and makes no further difference to the sound, right up until you clip. So, you have two constraints.
1.) noise floor. In reality this isn't much of a constraint since the self noise of a modern recording workstation is incredibly low. You have to really push it on the quiet side before his is even a problem.
2.) preamp/interface headroom. This is your biggest constraint. Most 'pro-sumer' grade stuff is decent but the converters get a bit non-linear when you really slam them. This matters because how a signal peaks (and the peaks of course are the part most impacted) really changes how it'll sit in a mix and how much clarity and perceived headroom your mix has, so you absolutely don't want to fuck around here, especially since you have nearly infinite perfectly clean gain in-the-box. You can always turn up a too-quiet signal, but once you hurt your peaks you're toast.
I usually track around -8-12db. This leaves me a ton of headroom and ensures he peaks are getting captured pristinely, and from a pragmatic standpoint also means that when I have 8-10 tracks playing back, I don't have to turn hem all down to keep the mix from clipping. And if you think about hat, that matters. Each track has only a finite amount of headroom, and if you're trackin hot enough that you hVe to turn all your tracks down so he master doesn't clip, you're just wasting some of hat headroom.