In this article that I have in my head, the argument is that continent divisions are first and foremost an educational tool, and that they should be defined along anthropocentric ideas. The splits aren't intended to speak great truths about the division of the world, but rather to divide the world into more manageable pieces.
Having "seven continents" when one of those divisions is totally arbitrary and leaves one side of the division with 60% of the world's population and 30% of the land area only reinforces Eurocentrism. Even at places like my kids' IB school, they still split things up along continental lines and all of Asia gets lumped together.
I started out fiddling with this trying to break it down on tectonic plates (partially because "how is Europe its own continent and not S. Asia?" is a central motivating question, but plates work for S. Asia but not globally.
So yeah continents are human constructs, not geological ones.
Also, what constitutes major faults vs minor faults gets confusing, and a lot of plate boundaries are pretty fuzzy and jagged.
Watershed boundaries are nice to divide on because they're already well known and well defined. They just have to be grouped together.
But there's also funny cases like the Brahamaputra River that you have to futz with. It drains into Ganges but only after cutting the deepest gorge in the world through the Himalayas, north of which it's the Tsangpo river. Do you consider southern Tibet part of the subcontinent because of that? Eh, maybe, but the agglomeration of like the 5 highest mountain ranges in the world seems like a decent spot for a continental boundary. So you just say, "well, it's the whole Bengal watershed except we're going to draw a line somewhere around where it makes a big bend around Mount Namcha Barwa (where the canyon is almost completely impassable) and call it a day."