There have been instances where something in the port configuration has an unexpected negative effect, but for starters let's look at how the network port(s) perform on their own.
What link speed do you get if you wire the 10G ports to one another (so no switch in between?) Does this change using shorter and longer cables?
Sorry for the delay, but now I've been able to test some more.
When connecting the ports on the server directly with the same cables I've been using to connect to the switch the ports come up in 10G immediately. I also got some new, presumably higher quality Cat 6A cables. These give the same behavior when used to connect to the Cisco switches, 100Mbit FDX. When connected in a loop between the ports on the server the new cables also produce an instant 10G connection.
Also, when messing with the cables, it came up as 1Gbps once. However, removing the cable and reattaching it produced 100Mb next time.
It feels like an issue with the auto-negotiation between the ethernet controller and the switch? However, I've been unable to find anything similar online.
The switches are:
Ethernet board: WS-X6748-GE-TX CEF720 48 port 10/100/1000mb Ethernet Rev. 2.4
Software: Version 12.2(33)SRE9, RELEASE SOFTWARE (fc2)
Software: Version 12.2(44)SE6, RELEASE SOFTWARE (fc1)