Do a little test with it. When the IRCAM is glitching and maxing out, try swapping to a different sampler mode a couple of times until it “behaves” itself. Would be interested to know if others could replicate this.
I have tried. I don't think there was any difference. I do find that after I have played a lot of notes, that then I have less trouble with the audio breaking up. But I have been assuming that that is because the work that IRCAM did has been buffered to some extent. If I play many notes in fast succession, or change chords quickly, it goes back to breaking up again.
I used the same sample in the regular stretch function. It worked perfectly without any trouble. But it also sounded no better than Kontakt's timemachine and timemachine pro. Whereas IRCAM stretch makes the sound better than the original sample. A bit like applying an Zynaptiq's Intensity. I demoed that recently and it uses an AI to analyse the sound and make adjustments that enhance aurally pleasing features of it.
Zynaptiq Intensity is very CPU intensive and the audio can break up sometimes when using it; IRCAM stretch seems to be combining similar processes to that plugin with time stretching. So I'm not surprised that it causes even more issues. In neither case does either my DAW (Studio One 5.4) or my computer register that they are overloaded. Again, that didn't suprise me with IRCAM stretch since it is operating on one core. It only needs to overtax that one core.
I tried a further experiment. The sound I had been using was a sample of a glass bowl resonating after being struck. I switched to a sample of a bowed string (a low tagelharpa string). IRCAM stretch started to break up when I played eight notes in a row as opposed to three or four.
So I would recommend, if you haven't already, trying different kinds of sample. It may just be that certain sounds take a lot more processing.
If your experiencing a worse case than I am despite having a more modern computer (i7 compared to i5) then that would be very odd. Either way, it is worth writing to UVI, as you plan on doing, to see what they have to say.
It would be very nice to have a batch processing feature with IRCAM stretch, to make getting the enhanced and stretched samples out and back in to another 'oscillator' more conveniently.