in this case I feel sure the unreliable component is the user.

Of course I know how to do it properly now, but it's a bit late.
I think I measured down to 4.7V at one point, which seems very low. I reckon the next thing to do is to take off all the 47uFs and re-do them with through-holes from the nearest solder points, rather than try to use SMTs where a track has lifted.....
Thanks for the advice, I think it's a low voltage problem. Would low ESR caps help?