when i first started heavily testing with sample loops a few months back, i had a lot of problems with 44/16. so i just run a batch conversion (reaper) to keep everything at 48/16. i always keep my mm set to 48.
i avoid sickocell/voxglitch samplers - their modules hav a lot of ram bloat. sickcell ones cut off your sample at 30seconds max. if you want to run longer samples, canard is the way for now. after doing a ton of tests to find the length limits before ram crashes, here are my numbers:
12x instances of canard in a patch: all 48/16 samples must be at 40s or below
8x: 70s or below
6x: 87s
4x: 95s
1x: 140s
those tests used patches that required only the following plugins to be loaded in at startup (about 13mb ram used at startup):
bidoo, fundamental, bog, mshack
weirdly the 8x gives the longest total run time: 560s/9.3m sample length in total.. but im finding that 6x instances works best overall w mm - 6 knobs on mm for trig+vol of each, and the 6 remaining knobs to control speed, etc.
anyway the only crash i’m getting now is related to the issue mentioned below (1st patch always loads fine, but 2nd patch after that might crash):
my remedy: when i need to switch to the next patch (and avoid the potential crash), i lift the mm out of the case a bit and hit that reset button then load the next patch. not pretty on stage, but works.