Time shown in the server list alternates between below 100ms and over 1000ms. Also there's some evidence of packet chew: broken blocks reappear and its more laggy than usual.

I've tried a router reboot in case the problem's at my end (wouldn't be the first time) but another server is fine. Could be an internet thing though somewhere between me and CF.

It doesn't seem to show up on traceroute, which is showing consistently low reply times and no drops.

Also vanilla server showing the same issue.

I'm guessing you know your stuff, so I'm going to explain.

High processing problems would not show on a network latency test via ICMP.

For future reference, testing the application layer instead of the network layer for connection problems would be your best bet, as the issue would be on layer 7, the application layer, instead of testing at layer 3 which your doing via ICMP, the network layer.

Although I will say working your way through the layers is the best way to find the problem

Probing the TCP/IP port directly would be a way to test the application layer.

Although this can't be 100% accurate due to multithreaded applications.




The cause:
Probably someone being silly with the pipes on tekkit again.


Edit:

The server list will give a more accurate measure than a ping or trace route as it queries the port directly, not via an ICMP packet. PSPing is a great tool for measuring latency in the application layer, http://technet.microsoft.com/en-us/sysinternals/jj729731.aspx