I’m scratching my head here, but the performance is less that half of that of 37. I’ve been hacking UNIX/Linux systems for 24+ years now, this one is a puzzler.
Thoughts?
Can you rule out a failing storage device (e.g., with smartmontools)? Check for common malware side effects that might go unnoticed on more capable hardware: high volumes of network traffic, connections open to strange endpoints, and high disk I/O. Use bpytop, htop, etc. to look for a rogue process using lots of cycles and accomplishing nothing. If that isn’t helpful, try the Phoronix raspberry pi4 benchmarks to find benchmarks results that are below par.
If the benchmarks are universally slow, compare compiler flags used in 37 and 38.