We just got in from Adelaide, SA and I decided to call up and find out what was going on with our case with Telstra's "congestion management department". Apparently the case is closed, can they reopen it? Well hold on, things certainly do seem better.
Upon running some tests now, we no longer appear to be losing packets periodically and the line speed is back up to ~5Mbps. I've been around the block a few times and I know that we can't really complain about the line speed because they don't guarantee it. I can deal with that, as long as it's consistent just about any amount of bandwidth can be managed in some way or another. It's the random packet loss and TCP connection failure that we can't manage, particularly with web apps that use a lot of small requests that fail with no discernible way to resubmit the request (here's looking at you, Blogger).
So what did they do? Asking that question was a recipe for bashing my head against the table. No matter how much I tried to impress upon the lady that I was at least decently technical, and if she just hit me with the technical jargon I'd grok most of it and anything I didn't understand now I could Google later, she kept on explaining it like I was five.
Apparently, they tore down and re-provisioned the virtual circuit between us and the DSLAM - at least that's about all I could understand after a three or four minute diatribe where she condescendingly explained virtual circuits versus physical circuits. Apparently, like rebooting a Windows machine when it's acting up, this fixes all sorts of problems and the issue wasn't actually congestion related at all.
So far so good, I'm back to tweaking bandwidth management again (mostly on the upstream this time, which is far easier to cope with) and we're only dropping a few requests here and there mostly because something is chewing up all the upstream on the WWW queue and I can't figure out what without firing up tcpdump, and I don't really feel like doing that tonight.