Yep, lots of hardware designs and products besides the software I have written.
PPC is just too confined being RISC (too many instructions required to be productive compared to CISC). When I was working on the Motorola coreless CPU project (basically a Xilinix clone with a microcode code engine that could be programmed), we got the CPU speed of the 68K emulation much faster than the 060 (much like Gunnar has done now with the Apollo core). It had the ability to swap endians and anything we wanted to do. Unfortunately, IBM really pushed the PPC architecture and that was the way Motorola went.
My 68040 core (with FPU and MMU) I wrote for FUSION-PC under Virtual Box for Windows runs circles around the Apollo core and WinUAE w/JIT on my 4930K Intel setup. Most benchmark programs break due to the speed. So, yes a CPU core running even on a low-end Intel or AMD CPU would make a better accelerator option. Powering it would be the problem though, so I don't see this as a real viable option without having a lot of extra power supply related hardware.