Good lead on that motorcycle coil.
Here is an update:
The engine in question is used to drive a pump for irrigation. The rough running (missing) happened on startup this spring after it had been sitting all winter. It turns out that the bad coil diagnosis was incorrect. I got the engine running smoothly by redoing the diagnosis, and finding the real culprit...
The problem manifest as a miss under load. The miss was intermittent to the ear (i.e. you can hear it). This was confirmed by using an inductive timing light as a spark indicator. There are six hv coils mounted on the top of the engine. It is easy to move the timing light inductive pick-up from one spark wire to the next. Initially, only cylinder #5 exhibited the occasional missing spark, which is where idea that the coil was breaking down internally came from. As I said, these coils are made from unobtainium, so getting a replacement is problematic.
After installing new plugs with no improvement, I repeated the timing light trick test, and this time I found that two of the cylinders (#1 and #5) were occasionally misfiring. I then swapped the coils on cyl#5 &6, thinking that the problem would follow a bad coil, but the misfire remained on cylinder #5, meaning that it was not the coil. The occasional misfire on cyl#1 also made me realize that two coil failures were very unlikely, so to go look at the only other thing where the problem could be: the magneto.
I took it apart, and sure enough, the magneto guts looked a lot like the innards of a normal Kettering automotive distributor. It has the rotor on the rotating shaft with the six cam lobes, the points that are opened by the high points on the cam, the "metal clad condensor" (capacitor), and different from an automotive distributor, the primary coil is inside the magneto.
Having owned several Volkswagen beetles as a youth, the problem was instantly obvious. The points were badly pitted, and were barely opening. I really need to replace them, but to get the engine running, I filed down the points to make two smooth, parallel faces, I reinstalled the points, set the gap to 0.025", and put the magneto back together.
The engine now runs smoothly as it did before. In hindsight, it turns out that the two cylinders that were misfiring are cam lobe neighbors, meaning that they are sequential in the firing order. This indicates a slight eccentricity in the cam rotation, so that side of the cam wasn't quite opening the points. Eventually, a third cylinder would have begun misfiring...
Sometimes you have to think about what you are doing when troubleshooting...