I have some experimental findings to report.
I found a plastic pipe that would fit over the verturi then I immersed the engine in a glass of kero and blew into the pipe. The first thing I saw was lots of bubbles from around the NVA nuts. Then I rotated the crank until the piston was just blocking the exhaust port. Blowing again, bubbles streamed from both exhaust stacks.
The cylinder on the Snipe does not screw into the crankcase. I just rests on a narrow shelf machined inside the crankcase. Screwing on the head holds the whole thing down with enough force in theory to make a seal. The bottom of my cylinder has some blemishes, either manufacturing faults or dents from being dropped. Its possible this is where the leak is. To test this I put a heap of grease around the sides and base of the cylinder and repeated the pressure test. No more bubbles from the exhaust stacks.
Its starting to make sense. The engine would start best when I gave a huge prime down the venturi. That would wet the cylinder base and make a bit of a seal. After about 2 seconds the engine would sag and usually stop. That was probably the fuel being burnt off and the leak returning. I'll bet with all that grease I could get a whole 10 seconds of running.
Now I have to figure out how to make it seal permanently. Araldite is looking attractive.