i disagree here - having a feedback loop that makes loud noise instead of cancelling is fundamental failure of their ANC algorithm. There's basically no protection mechanism incase of divergence.

If it's an algorithmic issue, can that not be changed?

my response was to the 'it should be an easy fix' sentiment. trying to address feedback loop without negatively affecting cancellation performance is extremely challenging, because for a mic/algorithm there's generally no way to determine if the error microphone is picking up a loud event in the environment or if the cancellation soundwave is incorrectly phased and instead of canceling its just adding more noise.