Why don't data warehouses, such as Snowflake with Snowplow, build their own reverse ETL tool, and can they take over the whole stack?

There's nothing really keeping them from doing that, but it's also not their core competency. You've built a data warehouse, you've built the platform on top of which all these things sit. If I were in their position, I'd just continue enabling other tooling to be built on top of my platform. Ultimately, Snowflake is still getting a cut of the reverse ETL tools, so it's not clear to me why it would be a high source of leverage of their time to do that as well.

There's also a really long tail of integrations that you'd have to build out and maintain if the data warehouse tool started to do reverse ETL. I would think more about, how do I enable other applications and other tooling to continue to be built on top of my platform? Like, innovation core to making the data warehouse itself operate faster, more efficiently, and more cheaply. You’re building the tools for this gold rush around data, so you don't have to be out there digging for gold yourself, because it’s your underlying tool that everyone actually uses to do that.