I think part of what made Hypercard so useful was that it made it really easy to make one specific kind of application, which at the time was 90% of what any single person could be expected to do by themselves. Most of the "multimedia experiences" people were making and buying were basically interactive slideshows, roughly as complex as a DVD menu. Hypercard made this easy, and you knew before you started whether or not it would be a good fit for your project. Under the covers, it was a Turning-complete language, but I'd imagine trying to make anything more complicated than Myst would be really painful. I suspect the developers of Myst probably ran into its limitations before they were done.

I think the main obstacle nowadays is deployment. Back in the day, you'd have Hypercard export a native application, and then copy it onto a disk that you passed around. Nowadays, the expectation is that people can run it in a browser, or download it from an app store. Web-based deployment means you have to embed it in a website using Javascript, while app stores require setting up a developer account, and building with complicated toolchains.

This isn't to say it definitively can't be done, just an explanation of why it hasn't yet.