but CTEs in SQL Server are not an optimization boundary like they are in Postgres. This makes a big difference in their proper use.

In SQL Server, aside from recursion, CTEs are purely syntactic sugar and thus a tool for organizing code. Whereas in Postgres, due to the optimization boundary, they are a tool for query tuning. If you rewrite a query in Postgres using CTEs just to clean it up, you are doing it wrong, since the CTEs will also affect the planner.