I remember marveling at the first time I saw multiple DOS prompts, all separate memory protected processes running on OS/2, knowing that rebooting the entire OS to handle a crashed program on 286 systems would be a thing of the past, knowing what a sham Windows 95 was compared to OS/2 and that NT was still a future milestone.
NT 3.5/3.51 was released prior to win95. It still had the old win3.11 look, but it had preemptive multitasking and protected virtual memory spaces.
I was excited about it as you were about os/2 apparently. And even as good as it's compatibility was it still didn't take off until winxp (3 versions later). So I think all the os/2 conspiracies need to be taken with a grain of salt when the slow NT adoption shows that os/2's biggest problem was probably corporate inertia.
Yep. For both OS/2 and NT, you had to buy an expensive high-end PC off a pretty narrow hardware-compatibility list, and the user experience wasn't too great (long boot times, laggy UI, etc.) The genius of Windows 95 was that it could sit on top of whatever strange legacy DOS drivers you were using and still provide some modicum of stability.
However, as the article mentioned, OS/2 did have huge corporate adoption, especially in finance and insurance. (Largely because IBM bundled it to their mainframe customers.) Most of these places switched to Windows when new OS/2 software dried up around 1996, but it wasn't uncommon to see an OS/2 terminal on a bankers desk well into the 2000s. (And I'm sure some places are still running it in a VM for some old internal software.)
Where IBM didn't have a clue is when they tried to market OS/2 to consumers and smaller businesses. Yes, you could go into the store and buy a boxed copy. But if you wanted to connect your computers to a network, you had to call an IBM enterprise sales rep. Assuming they even returned your call, and you figured out which alphabet soup packages to order, it was about $200/machine for the same basic filesharing which was included in Windows.
"The genius of Windows 95 was that it could sit on top of whatever strange legacy DOS drivers you were using and still provide some modicum of stability."
On the other hand, this helped Novell continue the DR-DOS lawsuit against MS. OS/2 1.x had little drivers initially, but it was beginning to improve by 1991, and OS/2 2.0 could use OS/2 1.x drivers.
You're correct that NT3.5 was out before Win95. That said, checking Wikipedia, OS/2 was out before either of them (April '87) whereas WinNT 3.5 was July '93 and Win95 was August '95.
Sidestepping chronology that is dating us massive, the productivity win of having minicomputer/workstation OS features available for microcomputers in both OS/2 and WinNT I just cannot overstate. Teh suck of rebooting a DOS machine multiple times a day just to do work is a memory best long forgotten.
Not to mention how Chicago (codename for Win95) was delayed and how MS was attacking OS/2 in the meantime (look up "Microsoft Munchkins" for one example). And it helped continue MS's legal troubles with Digital Research/Novell over DR-DOS. In contrast, OS/2 did not run on top of DOS at all, and DR-DOS 5.0 was released not long before Win3.0 was released. And MS released the OS/2 2.0 SDK to developers at the end of 1989. Yea, I am talking about the JDA break-up circa 1991. The MS-IBM JDA was not perfect, but the alternative MS chose was a lot worse and delayed native 32-bit programming by years (Intel released the 80386 in 1985!).
