Does your log mention database corruption?

No.

Include required information

Go is not cgroup aware. This means that Syncthing will be throttled hard in containerised environments with CPU limits.

The below screenshot shows a Syncthing container with a CPU limit of 1 core in Kubernetes. The first half of the time series is on a 13600k, with 20 threads. The second half is when the machine changed to use an EPYC 7763 with 128 threads.

This was sort of mentioned in another issue, but not really - #9357. Syncthing should probably use automaxprocs.