Each server and background process can write an associated trace file. When an internal error is detected by a process or user process, it dumps information about the error to its trace. This can be used for tuning the database.