First proper job was supporting OS/2 Warp 4 workstations on a Token ring network, talking SNA back to an IBM mainframe. So all this Windows using TCP/IP over Ethernet is something of a breeze.
We used to have a problem whereby every so often the entire floor we were sat on would just fall over - I remember standing in front of the Madge Networks "Ringswitches" and watching them go into full-on Christmas-tree mode. The relays used to make a great noise when they tripped. Cue many hours of playing hunt-the-knackered-balun under people's desks. One ring per floor, so when one device failed, 300+ people fell off the network. Nothing like a bit of segmentation.
I also remember another issue - encountered as they were upgrading to NT4. Turns out that the new IBM servers had been installed with dual NICs inside them. The primary was cabled to a dedicated 100Mbit HSTR ring, but the second was simply connected to the user ring on that floor, so when the HSTR switch failed (which was about once a month), then all of the servers would fail back to the 16Mbit floor ring, massively overloading it and reducing the entire floor to a standstill. And this was in the day when network management was very much an afterthought, so the usual diagnostic process involved staring at the box and hoping for divine inspiration!
In terms of tools:
1) Ethernet cable joiners - straight-through and XO
2) Assorted screwdrivers
5) RJ45 plugs and crimps
6) Leatherman (with belt-holster, because the ladies love a belt-holster)
7) Domain-admin password, written on post-it
8) Various network patch cords and fibres
10) Vanilla netbook purchased from PC World to get around USB-lockdown
11) USB-to-serial cable plus assorted console cables
12) Aftershave, just in case