I scale now the movement vectors before using them and removed the time_step multiplication for PL_MAXGRAVITY and PL_JUMPACC.
Yeah, this is a solution, too, although adding PL_GRAVITY per frame without scaling it with time_step onto pl.move.gravity.z is still wrong. On 15fps, time_step would be ~1,06 and on 60fps it would be ~0,26, so after one second, you get added 15*PL_GRAVITY = -5,25 for 15fps and 60*PL_GRAVITY = -21 for 60fps. That is why you have to multiply PL_GRAVITY with time_step before adding it.
