Some good progress in this area since last month's post.
I took off the wormwheel and cleaned thoroughly between the teeth with a glass-fibre brush, particularly what appeared to be the problem area. Still not smooth running, but an improvement.
More running-in and more lubrication continued to help.
Then I did what I should have done earlier - soldered the electrical leads to the motor terminals. With the wires supported to relieve pressure on the motor this is a very noticeable improvement. Perhaps the slight movements of the motor were causing momentary power loss from my loops of wire around the terminals.
Following up a comment on the previous post, I wired up my cheap multimeter to test the current. After some teething trouble (I'd never used it as an ammeter before), I found that the needle flickers at the "stiff" points. But this depends on the position of the motor - when it's unscrewed a quarter-turn, the flicker is more like a gentle sway. I'll have to devise a way of ensuring that the motor stays in this position once the body is fitted.
At the slowest speed without stalling, the driving wheel axles now make one revolution in over a minute in either direction.
Meanwhile I've bought a new controller - a Kent Panel Controls KFS/FC. It's fitted with a switch to turn feedback on or to a minimum. With feedback on minimum, at the lowest voltage the driving axles make one revolution in about 90 seconds, at a fairly constant 1.5mA. And at increasing speeds, the current rises to around 3-4mA.
Of course I still have to add wheels and coupling rods, but for now this seems as good as I can hope for.