While WebVR won't push the envelope of what's possible in terms of high end rendering, frameworks like [A-Frame](https://aframe.io/) will lower the friction of developing and distributing simple, more utilitarian VR experiences and of hopping in and out of them from your web browser or phone.
I can see many interesting business use cases that can be solved in a lightweight VR context, without needing a full gaming engine.
That's just it. Being able to render web content in VR makes it possible for your typical web developer to use their existing skills to build content for VR. Unity and Unreal are fantastic tools and very well suited for building games, but not everything is a game.
Besides this, instant access by url with no downloads and installs, no-app store "gatekeepers" and a free and open technology stack are solid reasons for why you would want to have or use WebVR.
And while performance definetly lags behind native, I'm not sure that it's going to be that far off. WebGL 2, Web assembly, WebGL in workers, will in combination with general improvements in graphics and rendering performance will get it close.
It was ahead of it's time. Mozilla's A-Frame is kind of it's spiritual successor. VR experiences starting with just a few lines of markup. Check it out https://aframe.io
VRML doesn't need a spiritual successor - it has an actual successor: X3D.
A-Frame looks a bit lightweight to do anything useful. You should have just implemented X3D.
Edit: maybe that's a bit harsh. But I really wonder why you went and created yet another declarative 3D format. X3D is an open standard that already exists, works well in the browser without plugins (X3DOM), and has fairly good support. It's also more advanced than A-Frame, and the syntax is almost as simple.
It's approachable and easy to use. That was the intent. But that doesn't mean its not powerful. It's built on top of a entity-component model (similar in many ways to that of game engines like Unity) that makes it highly extensible and customizable.
The declarative layer is an extremely thin abstraction. A-Frame has feature parity with three.js given its extensibility and entity-component pattern. I'd give it a deeper look.
I can see many interesting business use cases that can be solved in a lightweight VR context, without needing a full gaming engine.