abort process if too many select errors. Prevents error logs from
filling log partition.
after 90 select errors the log level will be increased for the next calls.
after 100 select errors the process aborts (will be restarted by systemd).
TODO: inspect sockets to find the bad one.