So, I've interacted with Kaldaien a good bit over the years and while he can have strongly stated opinions that are at times a little at odds with the intensity of an issue, it's not generally that he's fundamentally wrong. It's partially that intensity that seems to keep him so invested in the things he does.
Gabe Newell left a platform company that knew how important becoming a platform was. Microsoft has a great track record with software compatibility, if imperfect. Gabe didn't exactly want to create a platform, but there were very real difficulties in game distribution at the time and internet broadband had expanded. That Valve created Steam was not unwise.
That said, whether it's browsers, simple music playing, open source projects, software like Adobe Photoshop, etc there has been a rise in business and software complexity. Cross-platform compromises, unknown tool chains, unbuildable dependencies, dynamic ecosystems where a static program that just works no longer feels like a guarantee, megaprojects that increasingly solve a thousand problems you don't have.
Managing Steam is an unenviable task, depending on Chromium and Windows, expanding into an operating system project, hardware projects, even a bit of a social network... the mission has clearly expanded. The organizing principles of Steam have adjusted to a new environment and that has consequences. In the Steam UI you can go to View -> Small Mode to at least visualize how simple Steam could be, but all the other concerns still remain in the background.
When you have so many customers you can develop features for and deliver at scale, it's hard to reject the draw of shaping that clay. Hey, they've done a lot of great things with that power, but when you're so hooked in with momentum it's more difficult to undo the things you've done.
The problem is similar to browsers, though. When you become the leader, in a way you become the standard. You can still deliver DRM free games on Steam, but as far as I know you can't simply download them from the website the way you can with GoG. All the integrations Steam supports become a sort of standard that many customers expect rather than the true supportable platform baseline. Many games won't even release outside of Steam, so the only version that will exist is a Steam version.
Steam could address some of these concerns by developing a thinner optional client that aims for maximum platform compatibility and simplicity for use with a subset of games. That could then become another Steam feature that players demand or can filter the store by. Customers could opt to encourage developers to support that simpler client even if it means some features will be disabled.
Obviously, those games could get updated to need the fatter client, integrating more features, but that can be managed with good communication and options for developers or consumers. There can even be a different release channel or separate product ID. The simple client doesn't need an integrated Web UI either, since it could just open a URL in your browser that logs you in where you can buy games, check chat messages and so on.
I'm not sure if SteamOS could run on a Windows 98 era machine and run those games, but that would at least allow for a side argument that they maintain some compatibility. I kind of doubt that's high priority for them, though.