Hey Graham,
You've *definitely* got some pretty interesting looking results here - thanks for sharing these screenshots!
Judging from the looks of the timeline graph in "Fig 1." - I'd say that your inclination that this is a bandwidth issue seems to be on the right track.
The other servers and the switch showing up as hops in your route *is* a bit confusing to us as well, though (these kind of results aren't common at all). It could be that the switch is is attempting to route the packets that PingPlotter is sending out through other devices in the cluster (in an attempt to deal with any unresponsiveness/failures), but it's tough to tell for certain.
If you'd like, feel free to send over a pp2 file (or use the "File" -> "Share" option in PIngPlotter and shoot over the link) - as we'd love to take a closer look at this data! Or you can email us at support@pingman.com as well.
Best wishes,
-Gary