Google Docs may be an app, but a big part of the app, and confirmed to be one of the reasons for the migration here, is the part that renders the document to the screen. Only the sorts of techno-fetishists found on this site and in programmer circles would be able to make the argument you just have and not recognize the perverseness of what you're saying and what this decision by Google really means.
Documents are the quintessential use case for the Web, full stop, and they're the direct object of whatever the verb form of "Google Docs" would be. Google's decision here reflects the belief that, despite this, the Web is not suited for documents. We're not talking here about the mere act of the "app" chrome at the edges (surrounding the edited document itself) being switched over to use some more perfect framework better suited for e.g. painting interactive widgets, etc. No, it's right there in the announcement: "we’ll be migrating the underlying technical implementation of Docs from the current HTML-based rendering approach to a canvas-based approach".
Google is completely dropping the ball here and leading us down the path towards the picture painted in the top comment; what this is is just shy of an abrogation of their responsibility to act as a steward for the Web and the intention that it best serve users (rather than overpaid frontend developers trafficking in flavour-of-the-month fads, frameworks, etc and who already disproportionately receive attention under the status quo).
If the Docs team has identified deficiencies in the underlying standards-based model for—let's repeat it—presenting documents, then they are perfectly situated for translating that into feedback about how to improve things so that the Web is better fit for that use case. Even if that meant the Docs team going off into a corner, identifying what the problems with the Web are down to its fundamentals (DOM, etc), and then emerging with an entirely new approach for how to lay down bits so they might be better interpreted by the viewer running on the end user's computer, and then get the Chrome team to bake native support into Blink while disregarding every other vendor's possible objection to this act of steamrolling the standards process, then that would still be better in the long-term than what Google is doing here.