I feel so embarrased... but I think the problem of the drop outs etc is down to the powersupply.

To rule the PSU out I've just used the previously described PSU to power the 12v driver circuit, but used a 5v regulated switch mode wall PSU for the PIC. This was running at 5.12v. I had to common up the ground lines to get the loco to move, which also caused a bit of pulsing in the armature even with the pot at zero, but I couldn't get the lock up or drop out no matter what I tried.

So a re-think on the design of the PSU is in order.