A few for me that come to mind which haven't been covered yet.
*) Latency, jitter, etc when pinging a router means packets going
through the router suffer the same fate.
Never fails that I get a call about the latency changes that occur every
60 seconds, especially on software based routers. uh, huh.
*) admin/admin is okay in a private network behind a firewall
Oh, look, a console port!
*) Assign arbitrary MTUs in a layer 2 transport network based on exactly
what customers order.
*) MTU/packet/frame/ping size means the same thing on all vendors.
*) If Wireshark looks right, it must be right (unless Windows discarded
1 (and only 1) layer of 802.1q tags)
*) Upgrades should always be done, even when there's no relevant
security or functionality that is needed in newer code.
Amazing how many code changes break things which don't necessarily show
up in test environments but will show in production networks (Your mpls
worked for months with an invalid router-id configured, and then broke
when you change codes? DHCP worked fine, but after upgrade quit
accepting <300 byte DHCP packets?).
Jack