TechEmpower benchmarks, though, should be taken with a grain of salt. Many “framework” code are so aggressively tuned to the specific benchmark that noone will actually get that kind of performance in some real case. I’ve benchmarked many frameworks in their “default, out of the box” configuration, and I never saw such big differences.

See for example Data Update tab, pronghorn is performing not so good there.

For the pronghorn, I am more amazed if it is done by one person in the last half a year or so (since it is build ono coro-s) then by it's performance. It is too bare bones to be equally measured against "fuller" frameworks. But it looks like it definitely got some things right.