"Can it be done"?
Yes.
"Will it be done"?
No.
There are two options, implement the things in an FPGA, or have custom ASIC silicon made.
Custom ASIC silicon is far far too expensive for this to make sense. It could be done, reverse engineer the design, get new masks for a current process, but it won't happen. Why? It'll cost a few million US$, and it'll take a few years to do. No one in their right mind would spend millions of US$ on chips that in order to remain "custom chips" would opnly be allowed to be used in a couple thousand computers, adn taking a number of years for that many of these computers to be sold. There's no way it'd ever get paid for.
Using an FPGA instead of custom silicon begins to be practical. Even if you were to make a custom ASIC of these chips, you should do it in FPGA first for debug of your design, before wasting millions of US$ on debug of custom silicon. The FPGA chip is already built, all you have to do is figure out how big your circuit design is, how fast it must run, and pick an appropriate FPGA chip to hold it. FPGAs don't "emulate", they "are". There's gobs of generic circuitry in there, and your design selects which particular bits of curcuitry is actually used, and which interconnect pathways get used, and in the end you get a true implementation of your design, not an "emulation" of it. It's slower than custom siliicon due to the genericness, and there's effectively wasted space compared to custom silicon, but in our size market makes such things economically not absolutely throw you in a rubber room insane. There's a lot less things that need paid for this way, you don't need to buy masks to build custom silicon, you don't need to buy silicon design verification CAD/EDA programs, etc. but you will still need to buy verilog or VHDL simulators, timing verification CAD/EDA program, and pay engineers to do the work.
The FPGA way was attempted once. Remember the BoXeR? That's what that was. And we all know how well that worked out... If you wanted to use FPGA replacements in classic Amigas, you'll need to make adaptor boards sortof like the DKB Megachip board that fit into the motherboard socket, ,and have your FPGA on this board. I had begun the earliest thoughts of replacing Buster this way in my A4000T, hoping to get a speed improvement. Didn't get very far before finding other things to occupy my time. And if you don't have a socket on your motherboard, you'll have to put on there, surface mounted PLCC sockets can be found that will solder into place where your PLCC chips are soldered...
But again, this would be too expensive to make any business sense. It would take quite a while as well, as you still have to reverse-engineer the chips, considering the designs are lost. And even with the service manuals, hardware RKM, other various hardware books, and programming books describing the API for assembly metal-bangers, it will be very hard work and take a very long time to do properly.
Yes, it is technologically "possible" to do. But it's so far outside of anything even remotely resembling "practical", it will never happen. Anyone trying will most likely go bankrupt before finishing, just like anyone involved with the BoXeR project proved.
