For a long time, the 2.5 hz standard (18,000 vph) was state of the art. My guess is that below that rate there's a real drop-off in precision; I have never seen a watch, even a really old one, that used a slower rate.
The first move was a small one, to 2.75hz (19,800 vph). That must have made a lot of difference, as the heavyweights went with it: for example, it was included in Rolex's transition from caliber 1560 to caliber 1570; Patek's from caliber 12-600 to 27-460; IWC's caliber 85 to 852; etc. Presumably the manufacturers would have liked to go farther, say to 21,600 vph, but I think the required mainspring strength, manufacturing precision and lubricants were hard to come by. Thus, there aren't as many 21,600 vph movements around as one might expect.
Then there must have been some technological breakthroughs because suddenly everyone was manufacturing movements at 28,800 vph. (Not everyone; some companies really had a tough time getting over that hurdle and stayed behind for years or even decades.) I think this made a huge difference in watch performance in the low and middle ranges, although perhaps less at the top end. (It's interesting that some of the top brands were relatively slow to move to 28,800 vph. A really well-made 19,800 vph movement, with Breguet hairspring and variable inertia balance, is still quite competitive with a freshly- serviced high-beat movement and may gain an advantage a couple of years after servicing as it is less reliant on optimum lubrication conditions. In my limited experience, however, no 18,000 vph movement is in the same league.)
The next question would be why only a few companies moved on to 36,000 vph. As the owner of a couple of those contraptions, I think it's fair to say that the maintenance issues exceeded the timekeeping benefit. It's also interesting that Omega, for its coaxial escapements, scaled back from 28,800 to 25,200 vph. It's really quite a fine balancing act.
Generally, in movement design a faster beat correlates with a smaller balance. Since larger balances on the whole perform better than smaller ones, there are always tradeoffs.
In addition to purely mechanical issues, there is the question of how a movement's beat correlates with normal human activities. It's a bad thing to have the watch's natural rhythm tend to synchronize with the wearer's natural rhythm, as there will be a tendency for errors to be reinforced.
Leaving aside pure timekeeping performance, I do enjoy the look and sound of slower-beat movements.