It's hard for me to argue against this statement, but I can see a possible edge case. I think the "emulator" approach, if I may use that term loosely, might not match the timing behavior of real hardware. For people looking to do testing, tuning, and debugging on RISC-V, the gold standard is probably to do it on real hardware.
In the olden days we used to have quite a few systems with expansion cards that allowed running an alternate CPU in your PC, but reusing most of the parts like RAM, disks, graphics cards, peripherals etc. , because at the time they were so very expensive. With those you booted your PC into an alternate personality, when requried.
You had lots of microcomputer competitors e.g. Motorolas 68x00 on PCs, but IBM even sold tiny mainframes to go into your PC, long before Hercules (or Gene Amdahl) stared emulation or binary translation.
Some others basically ran another pysical system inside and in parallel to your PC, much like a single VM with an alternate ISA, heck even my Apple ][ with the CP/M Z-80 SoftCard did that.
I guess today cost parameters are just so different it makes very little sense, you can share test machines on the network and then there is always the cloud.
The only use case I have so far been able to come up with in a Framework notebook form factor at near enough economic scale would be military or security. For people putting their lives on line, a bit of masochism is simply part of the mission.