The gotcha is in the Windowed aggregate itself. It's a bit RBAR-like in that it seems to create a worktable with LOOP joins to the outer query, so its perf tends to suffer a lot with the "window" you're running this against is large. Meaning - if you were to need something like a "percentage of total", and your unsummarized data was large, it would likely be more efficient to use a derived table setup to create the "total", since you can control what method is used to link it into the main query.

On the other hand, if you need quick access to different aggregates (especially ones aggregated against different groupings), and the groups aren't huge, it tends to do VERY well.