I can see the argument for pushing levels with older equipment that lacks sufficient dynamic range headroom, or if the recordist doesn't want to do any level manipulation afterwards on a straight 2-channel stereo recording, only doing tracking and fading with no other editing.
With modern quality equipment and proper gain-staging I don't hear much if any difference between running high-ish levels and fortunately squeaking by without an over verses running comfortably low and adjusting levels later. Remember that it's easy to hear 'louder as better', so one really needs to very carefully make sure the final adjusted levels really are identical when auditioning such a comparison to make up one's own mind on the matter.
I do find the optimal recording level I aim for is somewhat dependant on the recorder and the particulars of it's metering. It's more a practical issue than a sonic one as long as the recording doesn't clip. I'm in the camp of recording low enough that I needn't keep and eye on the meters or worry about going over, and adjusting the level of the resulting files to whatever is most optimal later, since I'm going to do adjust it anyway.
If I have to raise the level so much afterwards that the noise floor of the equipment reaches or exceeds the ambient noise floor of the venue where the recording was made, then I recorded with too low a level and should have raised levels higher when recording, but that very rarely happens. Still, I'd rather have a bit of excess low-level hiss during the very quietest parts than hear audible clipping on the loud stuff.
Recording more than two channels-
I find those two arguments for running hot less persuasive when recording more than 2-channels. There isn't much gear which is capable of recording more than 2 channels which will have critically limited dynamic range capability. And unless recording multi-channel surround, you're going to mix the resulting channels down to 2-channel stereo someway or another, so shooting for an initial recording level which is high enough for a final output file as a way to avoid further level correction is pointless.
For me the more interesting thing and more persuasive practical argument becomes maintaining the same level relationship between all recorded channels throughout the entire recording.. or throughout all related recording in a series - such as all recordings made over the course of a weekend at the same music festival. I don't want to find myself having to lower the gain on one channel or pair of channels because it's in danger of clipping, while leaving the other channels untouched. I'd then want to make a note of the time of the adjustement, to see if I needed to compensate for it afterwards in my mix, which would be a PITA. To not do that would mean that whatever mix I dialed in as correct at the beginning, would get thrown out of whack once the recording got to that point. I'd rather find a good relationship between all recorded channels and adjust them all together simultaneously when necessary (say at the start of each act if necessary, like for a loud electric act taking the stage after a quiet acoustic one), to keep my mix relationships the same throughout, making the mix far more easily manageable than if just one or a few channels suddenly change levels while the others remain the same.
With 4 channels it's still not too difficult to quickly turn all four input level knobs by the same amount to make those across-the-board recording level changes, but it's much easier when using a recorder which can 'gang' the input level controls together across all channels. Then, one knob turn adjusts the recording level of all channels simultaneously and by the same amount. When recording more than 4 channels it quickly becomes far more complicated, and such channel 'ganging' of input level controls becomes even more valuable.
So when I'm recording 6 channels on the DR-680, I'll first dial in the relative level relationships to where they need to be, by adjusting individual channels and sometimes by temporarily ganging stereo-pairs, or whatever. Then once the relative relationships are good enough, I'll gang all 6 channels together so that everything remains balanced when I need to make a gross level change. Like rocksuitcase, when recording to the DR-680 I find shooting for peaking at the -16 line on that recorder's relatively small meter a safe bet for most material.