Pitch variation at start of (audio?) file

When starting the playback of an audio file, VLC reports:

core debug: Buffering 96%
core debug: Buffering 99%
core debug: Stream buffering done (1018 ms in 53 ms)
core debug: Decoder wait done in 0 ms
core warning: playback too late (84934): up-sampling

And then after some 15 seconds:

core debug: resampling stopped (drift: -11878 us)

So why does VLC find it necessary to resample in the beginning ? Althought many people apparently can't hear this, it sounds terrible to me. To my nerves it has a similar effect as scratching a blackboard with your nails...

Why does VLC report: "playback too late"? Too late for its own intentions? That should be no problem, I'm fine with 80ms longer waiting ;)

When there is a discrepancy between a and v sampling this will not become clear in the first few seconds, but only after some time, when a shift starts to appear. Then a tiny amount of pitch changing (less than 0.1%) by means of resampling may be needed. Resampling only at the beginning should never be needed. If there's an offset between a and v in the beginning, the missing part of a or v should be filled with black/silence. And for audio-only files there is no reason at all.

To my knowlegde this initial pitch changing has been in VLC for years.

I really hope someone can find this problem.