That said, I will concede that there is definitely an audience for low-power devices obviously, and maybe FPGAs will fill that niche.
But that also doesn't necessarily offset the latency of getting inputs into the emulator in the first place or in getting video frames from the emulator to the display.
That's not an inherent limitation of software emulation, no, but for all practical purposes it is as nobody is doing specialized software emulation for a specific set of hardware to bypass the normal OS/graphics stack. It's instead "throw linux on it, and fullscreen an off the shelf emulator"
I agree with most of your point here, bus is that particular one true? Doesn't the Gamecube (for example) use a PowerPC chip? I would be very surprised if isn't an effort to make FPGA implementations of the PowerPC.
I realize that there's a lot more to making a clone system than "just recreating the CPU", but why would making an FPGA for something after the N64 be impractical?
I think this might be more important as time goes on as well. Getting cycle-accurate emulation of anything more complex than the Dreamcast I think will become pretty prohibitive to do in-software. I suspect something like CEN64 might end up being the last cycle-accurate emulator out there (though I would absolutely love to be wrong on that). The advantages of FPGAs would make themselves substantially more apparent at that point than the Gameboy or NES.