NeXT boxes could not compete in the visualization space against SGI and they lagged behind Sun, Digital and HP in price/performance, even after ditching the MO disk for conventional ones. It also didn't help much they had a weird Unix OS that didn't have X.
They were well rounded machines and their development tools were excellent, but that's it. As much as I would have loved to have one, I could never justify the tradeoffs between it and, say, a Sun pizza box.
The good news is that in the talk, Steve mentioned that the Enterprise WebObjects (probably the one you needed: EOF was for WebObjects what Cocoa was for OS X) license was $25k so it had started to drop. Even back then free was tough to compete with...
At the time, even though it was easy to see that free software would eventually reach a critical mass and overtake all the system software space, the market was not nearly there.
While I used a lot of Perl (which was free) to write dynamic pages, all my software was running on Solaris and IRIX on very expensive hardware. HTTP was served mostly by Netscape's proprietary server with some traffic on NCSA HTTPd. Linux, IIRC, had terrible SMP support back then and Windows was a joke.
Put differently, any platform needs a critical mass to survive, and if they price it so high, it won't have that critical mass, causing it to die, in which case the development cost would be wasted.
Surely people in the 80s weren't so dumb as to not understand network effects, so what am I missing?
My MacBook would like to have a word with you. ;-)
NeXTSTEP was only dead to the market - it was still a pretty good OS, specially when compared to MacOS 9 and Windows. NeXT failed against Sun and SGI, not against the PC market. They never even tried that.
That's 2019 thinking. In 1990s almost no one would think that, especially in the world of commercial software.