Actually the PC market killed Tandy in more ways than one. If you look at "Background on LSI vs Tandy and the creation of 6.3.0. " in
http://nemesis.lonestar.org/computers/tandy/software/os/logical_systems/lsdos6/src631/history.html a follow on implication from actions there is that enough people in key positions in Tandy had lost interest in the non-pc clone stuff.
Saying that they 'lost interest' is probably being too nice; they were actively blocking things for the non-clone stuff (search comp.sys.tandy for the threads on the 6000's MMU that allowed access past 1MB).
...the 6000 could have been turned into something to match the SUN 3 and Apollos of the late 80s (which made quite a bit of money thank you very much).
Hmmm, up to a point I would agree. Getting a 68020 on the limited size card that the 6000's card cage can support would have been tricky. I have here a 68020 CPU card from an '80's Proteon ethernet router, and it takes a full Multibus card tightly packed (but that's with RAM).
In terms of raw performance with text-mode apps, the 'Super' 6000 with an '020 might have done pretty well; I/O is still going to be a bit hobbled, though, unless a Z80 upgrade were to be in the works. If you were to upgrade the Z80 to a fast chip and/or a Hitachi HD64180 at 6MHz or higher and allow the '180 to have a much larger window into the '020's RAM space you might be able to remove the bottleneck. The similar spec'd AT&T UnixPC/PC 7300/3B1 (68010 at 10MHz as I recall) was quite a bit faster with the '010 doing the I/O with no Z80 running that portion. I do remember doing the '010 modification to a 6000, but I never finished the 10MHz upgrade, although it was possible with some chip swaps and faster RAM. (I didn't finish because I bought a 3B1 at a good price and let the 6000 languish a bit.)
The biggest technical problem the 6000 had versus Sun, Apollo, etc, was the lack of a graphical framebuffer. Apollos started with monochrome screens (Sun's did, too), so color would not have been a hard requirement. The Unix workstation business was highly graphical in nature; even the 3B1 has a reasonable framebuffer and a usable non-X11 GUI. The next problem was hard disk performance.
The biggest total problem preventing the 'Super 6000' from coming to fruition was the PC clone business. Tandy likely did more volume in the 3000 and 4000 series than Sun and Apollo combined did with 68K-based workstations.
I base that last assertion on a document I have from the collection of an employee from a local factory. This local factory was part of a much larger company which standardized on Tandy/Grid 3000 and 4000 series PC's in the late late 80's and early 90's; this is a huge company, and they went enterprise-wide to Tandy/Grid PC's; I am reading a copy that I have of their recommendation document from February 1990, and their 'corporate IT recommended PC systems' were, in order: 1: )Apple Macintosh family; 2.) IBM PS/2; 3.) Tandy 4000 and GRiD 386sx products. The document is marked company classified, so I am not at liberty to say what company or to provide copies. (Yes, this company was large enough that it had a four-tier classification system; this is not 'government classified' information, though.)
I know that there was several hundred Tandy 4000 series machines in the one factory, and, again, this is a company in the top 20 of the Fortune 500 with worldwide employment in the hundreds of thousands.
A 'Super 6000' would have taken technical people away from the much more lucrative 4000 business.