Rebooting a device to solve a problem is good.
Rebooting a device to then wonder what was the issue is bad :-(
Some troubleshooting would have been useful.
Could you ping the service at the time of the problem ?
What kind of keepalive ? TCP ? HTTP ?
What CSS software version ?
How is the server connected to the CSS ? directly or switch ?
Did you verify L2 ?
Did you capture a sniffer trace to see if the CSS were sending keepalive ?
Thanks,
Gilles.