Generously contributed by former AmigaOS team member, Peter Cherna
When I started at Commodore in the Amiga software group in mid-1989, I got to see the then-current state of the OS, which was approaching beta on 1.4. So much good stuff, and I thought "this should be called 2.0". I was told that "No, 2.0 is the release where we get to break everybody", in other words the hypothetical release where "doing it right" was more important than "making it compatible".
Anyway, we had a beta target date of mid-December 1989. By that time we had some of the new Preferences editors up, and the GUI style of the gadget toolkit was taking form, but the window borders were pretty much as ever. One Monday we came in and two of the engineers (Dave Berezowski, who had primary responsibility for the Workbench application, and Steve Beats, who was doing file systems, drivers, and a lot besides) were looking mighty smug, but wouldn't say why. After the whole team straggled in, they showed the first cut of the "new look"—mostly revamped window borders. Their weekend effort made it clear that we could and we should adopt a new look more broadly, which led (among other things) to:
At that point, with the striking visual changes, it was pretty obvious to everyone that we had to call it 2.0 and set aside the notion of saving 2.0 for the "big one".
So really, 2.0 was the final name for the release that was going to be 1.4, but once we had the new look, the new number was just a matter of time.
The A3000 was launched with 2.0, which contained one quite serious bug (I forget what) that was immediately fixed in 2.01. (That was about all 2.01 added).
2.02 was a huge effort internally to fix the worst problems and get stability way up. A lot of people labored incredibly hard to make 2.02 quite solid. Bryce Nesbitt managed the core OS subgroup and did an amazing job at making a very disciplined release. One or two loose screws meant we still needed a 2.03 to fix that up, and 2.04 was another big effort and was very stable and compatible with a huge number of applications that for various reasons good and otherwise did things that were not strictly by the book.
There are a lot of unsung heroes from that period who poured incredible efforts into 2.0x stability. Some like Bryce, and Mike Sinz, worked for Commodore. But we had many external testers, most importantly Bill Hawes (who wrote ARexx), who did an incredible amount of stress-testing of the software, all of which not only made 2.04 extremely stable, but established a culture of stability that lasted as long as I was there (through 3.0), and presumably longer.