This PR handles two processes causes

1. Where a TCP session prevents one child to exit quickly, causing the others to become because of the blocking read waiting for QUIT_SYNC_ACK that old-main sends when all the old serve children have exited.
2. When processes take long to exit they exit after the "reload" process becomes "main" again. In server_main(), the loop starts with reaping exited children, and then a poll or ppoll starts waiting for 60 seconds. If old-main or old server childs took long to exit, they will exit during the 60 seconds poll en become . By changing the wait to 1 second, such processes are reaped at least once a second.