A friend of mine started a company in 1983 to develop Mac software. They didn't want to go the Lisa route for a couple of reasons. First, they wanted to write in C, not Pascal. Second, they wanted several developers working simultaneously and could not afford to get a Lisa for each.
They knew that MacFORTH was going to be available pretty much concurrently with the release of the Mac. They knew they could use that to write programs on the Mac to explore and reverse engineer things.
They bought a Callan Data Systems Unistar 200 Unix workstation [1], which was a 68k based all-in-one non-graphical Unix workstation. They also got a bunch of serial terminals (ADM3a's, if I recall correctly) and hooked them up to the Callan, and gave each developer a Unix account on the workstation so that several developers could be working at once.
The idea was to write in C on the Callan, use its compiler and assembler, and then somehow turn that object code into code that would run on the Mac.
They had the Callan connected to the Mac over a serial port, and a program in MacFORTH that would accept object code from the Callan and invoke it. With this setup, their developers could start working on parts of the product, with test harnesses in MacFORTH to let them test on the Mac.
My memory is getting fuzzy, but I think somewhere in there they wrote a Mac linker on the Callan that could generate Mac executable files, which they could send to MacFORTH on the Mac and have it write them to floppy.
[1] I was a systems programmer at Callan, and was able to get my friend a nice discount. I think it came to $5000 for a nicely configured workstation.