Possible stability issue in variable filter at low Q and high cut off frequency.
https://forum.pjrc.com/threads/27896-Teensy-3-1-Audio-Library-limits?p=68051&viewfull=1#post68051
Looking for integer overflows, I modifed by guessing the limit on fmult as imposed by https://github.com/PaulStoffregen/Audio/blob/f4bdcb486212cd20366f1d52c17ad94a93022555/filter_variable.cpp#L130 .
Lowering this limit by a factor of 0.8 made the filter stable in the full parameter range for frequency, Q, octave range and control input as possible by my application:
if (fmult > 4302623) fmult = 4302623;
Where does the original number come from? How it is computed?
If there is any useful rationale about it, I would set a new feasable limit and issue a pull request.