It may be worth grabbing a copy of the /proc/sys on a clean boot and then a copy after the sysctl changes so we can get and idea of any specific tweaks that may have occurred.
Just to note, I've been running the stress-ng command as noted in comment #3 on a 24 CPU ARM64 Synquacer box with the alleged faulty kernel and cannot reproduce the issue.
It may be worth grabbing a copy of the /proc/sys on a clean boot and then a copy after the sysctl changes so we can get and idea of any specific tweaks that may have occurred.
Just to note, I've been running the stress-ng command as noted in comment #3 on a 24 CPU ARM64 Synquacer box with the alleged faulty kernel and cannot reproduce the issue.