I thought it was driver problems at first because it happened so infrequently, but it has gotten worse this year.
It took ages to narrow it down to memory faults since I’ve been running stock settings forever. Stumbled upon this tool: https://github.com/GpuZelenograd/memtest_vulkan/
Found a load of errors, which went away after the down clock and it’s been stable since. It must be just age related degradation, temps were never high but I know the repeated temp changes have an impact.
I bumped the power up on mine, I might be wrong but I assumed more power but less speed would allow it to keep stable and it seems to be working so far.
https://github.com/GpuZelenograd/memtest_vulkan/ - was the tool I used to confirm the memory issues. Just need it to hold on till the 50xx series!