As a software engineer with 25 years of experience, I know how this stuff is built. Assuming that the playback engine is there to decode an audio file and then transport to the DAC, via direct I/O pathways, without modifying the PCM data, I am puzzled by what could be causing perceived differences in sound. Not the same thing as saying: "...it's a mystery and so must be snake-oil."
My apologies. Most of us think that part of the reason is activity within the CPU, I/O et cetera which, in turn, subtly modulates the supply rails, plus alters the resulting eye pattern and recovered noise floor. Noise directly radiated from the circuitry as duty cycles and other activities change may also be an influence…dunno. Alas, that what research scientists (and grad students) are for! ;-) I know quite a few senior audio-only software engineers, and they freely admit that, at times, they too are puzzled by the subjective change heard from seemingly minor changes in their code.
Caution: anecdote not directly associated with the above topic…
I was reading a really interesting interview with Bob Ludwig, of Gateway Mastering fame, in a recent TapeOp magazine, and he mentioned how bad early PCM DACs and ADCs were, since they were designs "ripped from the textbook.” With Matlab, almost anyone can design a signal processing block, but it’ll most likely sound like crap. It usually takes experience to know what form or approach will sound good, maybe without knowing why. The “why” will come later, as the state of the art is advanced.