You will probably only notice it when you bounce your tunes as most DAW's have some headroom over 0db to accomodate high levels.
Should hear it on playback. The floating point internal numbers have to be crammed into the fixed-point pipe on the way to your DAC.
At mix time, the important question is not so much where the peaks are hitting, or even really where the average sits, but in the dynamic range - the difference between those two values. If your average is like -7dbfs, and your peaks are at +2, then it's just too loud anyway and you might as well turn it down. If it averages -18, though, you're probably going to have to smash those peaks down a bit somewhere along the line.
Usually when you look at a waveform of an unmastered mix file, you'll see a row of a lot ofpeaks that's pretty consistent at least within sections of the song. Poking out above that row is some less frequent, but still pretty regular peaks that go higher (quite often kick or snare hits) and then above all of those there will be just a very few spikes that poke out several db above even that. These I call "abberent peaks" because they really are just accidentally everything in the mix is actually pushing in the same direction at the same time. They're usually very short and very rare, and you can almost always just clip them off without anybody knowing the difference. I prefer a bit more gentle curve there, but it can work.
But if you just render it to fixed point as is, your peaks would be right at 0dbfs, which is also not really best practice. Use a limiter or clipper to chop them off somewhere below 0.
I dont worry too much about those abberent peaks in mixing, and lately I've been rendering my mixes to floating point waves, so that I just put the master fader at 0 and worry about peak levels at mastering time.
