The web and the way MS handle the web killed it. Microsoft was pushing everyone towards these horrible activx components. Many moved to PHP or Adobe Flash.
I was using MS Access back in the day to solve business problems - it was like a VB6 DSL focused fully on data-driven applications. It was extremely time efficient thanks to that focus - what took me days to build took a VB developer (or PHP developer) months. That died thanks to web (and the web version being horrific and a dataleak waiting to happen).
The move to C# and the path they've followed since then was, looking back, the way to go. I only wish that they had done more to make the experience for native programmers better out of the box.
In my experience, ActiveX was not what they pushed. They pushed Asp.Net Webforms which used VBScript for the template language. The ViewState monstrosity made it an awful experience. Specifically how it send way more data than needed to and from the server on each request. I shiver when I think of it.
The real problem was worse than that: it was not a technical, but a legal one: Licensing!
You need to buy CALs for every concurrent user (yes, not only SMB/terminal services users) was the way to go according to some MS license experts, and the licensing was totally incomprehensible, as is today. No wonder most didn't want to risk building a business on these maybe not so solid foundations.
Still I only dare using asp.net for serving public webpages from Linux.
Of course the technical problems were also present, but many didn't even get to think of that as the BSA bullying was keeping them away already.
I think GP is describing the period before the .net switch. I remember back in those days having to do web stuff on MS platforms, which meant classic ASP, which was great for the job it was designed for: gluing together ActiveX components.
Unfortunately, nobody anywhere ever trusted ActiveX because it (at least gave everybody the impression it) had holes the size of a bus, so you were stuck using a limited language with no library support to do all the glue and all the heavy lifting, it was a very unpleasant period of time if you found your way into a Microsoft shop.
When the web was new, it was so clearly not designed for apps that pretty much everyone agreed it was absurd to use it that way. The UX and DX was just such a horrifically huge downgrade. Instead, HTML was for documents and linking, and apps would be given a rectangle (maybe the entire rectangle) through which they could appear in the browser. Netscape partnered with Sun for Java applets. ActiveX was therefore a response to that.
ActiveX made a lot of sense for Microsoft at the time. Firstly, it let the enormous base of Windows devs put stuff into a web page almost overnight. Secondly, it leveraged existing tech that was already designed for embedding one app's UI inside another app (OLE), and for shipping widget libraries to visual designers (OCX). Thirdly, ActiveX was a lot faster and more efficient than Java applets, being based around tightly written Win32 code. Finally, people weren't so focused on security back then and anyway, Microsoft had recently been investing in adding code signing to Windows. It appeared to them at the time that code signing would be sufficient to stop abuse. We know in retrospect it didn't quite work out that way for them, although it's been pretty successful on Macs.
And in fact ActiveX did see quite wide usage, albeit mostly as IE's answer to Netscape's plugin interface. But it didn't take off in a big way because:
1. ActiveX apps were native code and native code is very low level and thus very large compared to JavaScript and HTML. Performance is great because the CPU eats it directly with no intermediate translation layers, but you pay for it in code size. Back then bandwidth and CPU were both tight, but HTML/JS were light on both at the cost of UI latency and primitive controls. People picked low startup time followed by high latency and primitive controls.
2. The UI they picked for code signing was a (standard at the time) popup modal dialog. It had very poor usability, being filled with tons of complicated words, often asking you to approve software from a company you'd never heard of (i.e. products/websites did not use the same name as the company that produced them) and worse, malicious web pages could put you into an infinite loop where if you clicked cancel they'd just immediately ask you again, forcing you to click yes. Which was equivalent to downloading and running an unsandboxed program. Back then browsers went years between updates so once they shipped that mistake, it was very hard to fix quickly and so the reputation of ActiveX became very poor.
3. They had no way to sandbox native code at this time, the Windows kernel just didn't support it at all and it probably wouldn't have occurred to them anyway.
The first problem was fundamental to the approach. The second problem was an unforced error.
They could have tried to switch over to a sandboxable bytecode; VB used what they called "P-code" on an interpreter already. But they didn't. Java was becoming too popular, too fast, and they lost confidence in their own tech stack. Instead Gates threw the company into a crash effort to clone Java. By the time .NET had the ability to do ActiveX controls they'd already moved on from that approach and were trying to make DHTML work instead.
I have the opposite take, as a non-native programmer in the early 00s.
All the PostBack loops in ASP.Net and WYSIWYG editing in VS.Net 2003+ felt like they were trying to appease native programmers to a fault. VB.Net wasn't bad but it was frustrating how hard it was just to script something without boilerplate. Being limited to IIS was also bunk.
I was using MS Access back in the day to solve business problems - it was like a VB6 DSL focused fully on data-driven applications. It was extremely time efficient thanks to that focus - what took me days to build took a VB developer (or PHP developer) months. That died thanks to web (and the web version being horrific and a dataleak waiting to happen).
The move to C# and the path they've followed since then was, looking back, the way to go. I only wish that they had done more to make the experience for native programmers better out of the box.