The trouble with cyrotrons was that you could make them fast, but you couldn't make them small. Not by IC standards, anyway. They have a magnetic component, which limits the packing density and makes fabrication by lithography very tough. So mainstream IC technology won out.
In the early days of computing, the military was way ahead of the civilian market, because the military was spending the money. In the early 1980s, the civilian market caught up, and by the late 1980s, the now much larger civilian market was ahead.
What did they do with it?
Take a look at  and read the description of the custom hardware they had added to HARVEST, their IBM 7030 STRETCH computer. That gives a sense of the hardware wanted for statistical cryptanalytic key testing.
One surprised me by using mercury. Proving no buffer overflows exist in the code was rarely more significant far as hardware itself went. ;)
When they were describing his research in the beginning of the article, I thought "I wonder if he worked at Lincoln Lab, this is just the kind of thing they'd be into". And sure enough he did work there later on.
Also cool to see that letterhead from Lincoln Lab, Division 6 (solid state research) which was around way back then! FYI, I worked in Division 8 at LL back in the late 90's (doing electro-optics). Awesome place.