'Isn't everything' is a huge understatement. It's a performance nice smoke test for a CPU in that if it isn't hitting the numbers you'd expect given the microarchitecture (issue width etc) you know there's a problem and due to the simplicity of the benchmark it shouldn't be hard to analyse and track down.

Though there's way more a modern high performance CPU needs to do. It doesn't do anything meaningful to stress the memory system for instance. Hitting good Drystone numbers is stage 1 in a long process to building a modern high performance CPU.

That's fair. Perhaps "is only an imperfect early indicator" would have been better than "isn't everything".

As I understand that is performance in a simulator, and it is unknown what delays are going to be between pipeline stages on real hardware.

Performance in a simulator is cycle accurate to the real world.

"delays" affect the MHz, not the performance per MHz. Tools for predicting MHz are fairly accurate.

Also, VROOM works in FPGA not only in verilator.