Gave up for a while on finding the reason the CMTS wasn't properly doing a reset. Instead, I concentrated on finishing the feature I was working on previously; namely recording history for various counters. (Should there be a comma after namely? Looks like it, but then the sentence reads awkwardly.)

That feature took a surprisingly long time, mostly because I didn't realise that a refactoring error was spawning 5 new recorder "threads" for every recorder on each iteration. That resulted in these weird errors where the same record would get added twice, but as far as a given recorder was concerned, it was only processed once.

Once I tracked that down, everything worked as planned. It's a pretty heavy feature, though, and at the moment I haven't added a web or xmlrpc front-end yet. Have to do that next I suppose.