I "solved" the problem by using Maelstrom's implementation from the same thread. Whatever differences there may be in the sound are in no way greater than the fact that the synth actually works now, as I'm performing this evening! :)

However, as this is an interesting issue possibly indicating some sort of regression between Pd 0.38 and 0.48, I've attached the relevant part of the patch. I'm not very good with nuts and bolts DSP stuff, so I barely understand what's going on, I'm just a good black-box programmer who knows how to connect things up based on allegations. ;-)