"Slowdown in the audio" suggests codec issues[1], which might or might not explain the linked crackling too. Or the glitch that somehow interferes with the speed tracker also causes 'normal' sound quality issues.
I rarely see audio go out of sync with video (if that's what I understood you to say) although sometimes I can be a bit rough with ffplay's skip on/back functionality in a high-def video stream on a low-end machine and the video is blank for a while, then 'catches' and fast-forwards a bit until it resyncs with the audio that has been playing just fine until this time. It's fun to lip-read characters (in a conversational scene with no othe cues) and try to relate what their lips are quickly saying with what I heard said a short time earlier. (Best when the desynch is at a point where a man is mouthing different words to the woman whose voice is active (or vice-versa), at a slightly different tempo, yet it
still imaginably matches...
)
Not sure how much of this helps, except for anecdotal proof that "it happens".
Out of interest, if you covert the files to different formats (if you can) from your habitually cobsumed kind, does the issue happen just as much? That means not just changing the container format, but the layer formats within, so (Ogg) Vorbis audio compared to (MP4) MPEG Audio Layer 2, or whatever applies. Might not sort the problem, might not even be practical to test, but could rule out my idle thoughts if I'm totally off track.
[1] I've got a machine that will occasionally go a bit gaga with audio (this being VLC, as the player, not ffplay as I describe I use elsewhere). As if it missed a new fourier definition or perhaps enacted one it shouldn't have in the 'pallete' of frequencies it has to service by index, everything is suddenly noticably (but must have been shifting for a while) at a lower frequency - it plays mostly spoken word, not music, so it isn't immediately obvious. It'll gradually depen and deepen further until it sounds like a slowed-down tape (but still at 1 second/second playback, so not 'drawlingly' slowed down) which then resets on starting the next cued-up MPwhatever file. Which makes me think it's a memory glitch somehow shifting the 'frequencies needed' table each time it passes a given space-saving 'delta' difference. Replaying the file later will not show the same issue, but travelling the "progress slider" back to a timestamp before the issue was noticable and it is as bad (and continues to distort worse and worse), so it's definitely not the file data itself but something spontaneous in memory (whether main RAM or audio chip cache). But I live with it, as it's only once every thousand or (a lot?) more playings that do this.