One thing that hasn't been talked about much is that P4 kind of gave you lock-in to Tofino given that it was the only switch chip to implement it. Now those concerns have become real with Tofino being canceled. In theory Cisco Silicon One also implements P4 but it's probably not the same dialect and many people don't want to pay Cisco prices.
I also think P4 has been held back by the fact that hyperscalers mostly converged on a "dumb underlay" architecture where they don't expect the switches to provide any interesting features. Then P4 and Tofino came along and P4 has little benefit in that architecture (except maybe QoS which seems to be what Oxide is doing).
P4 may end up being used more in DPUs where hopefully we'll have support from 3-4 different vendors (with different dialects of course).