Maybe I'm not following the author's train of thought completely here but I don't see anything at all here that adds up to his conclusion.
There's a single example given around a very particular use case around importing packages (for which it turns out there's a pretty simple solution, if one wants to keep the same approach).
Suppose that this particular choice for Deno is not your favourite idea, there are a lot of improvements everywhere else.
Hard to predict the future but my hunch (since that's all anyone has at this point) is that yes, it will eventually "replace" Node, for some value of "replace".
Mostly agreed... I think of deno more as an alternative major change (much like python 2/3 split). Not sure if deno will take over, there are a lot of places that should be shorn up... one of the big ones for me is a consistent ffi library, and while node-ffi package works, it's not exactly transparent and imo should be in the box.
Deno is pretty solid in most other areas though... only native modules need a bit more considerations for the use cases without a good alternative. That's a little outside my comfort zone as a developer, but there's work in progress for all of this.