Normally I would say that is due to one of these settings:
Code:
wait_timeout
max_allowed_packet
However, you've already changed those.
It could be due to a memory limitation (although swap space should be available - but it may not be).
You may need to monitor the server processes to see exactly what's happening when the install routing is running.
The server error log shows a lot of stuff related to MySQL being killed, but not necessarily why.