That's an interesting and seemingly obvious approach, so I'm curious why it's not really discussed.

I'm guessing that at current wind/solar capacity, any fluctuations can be absorbed by just slowing down the flow out of the reservoir, rather than ever having it reverse.

You also need a downhill source of water, which means any dam that's the last one downstream can't really use this approach.

You could maybe even get away with an entirely closed-loop system, but if you look at the scale of the problem, you need either a huge pair of reservoirs (lake-sized) or a huge elevation difference (mountain-sized), or preferably both. It doesn't appear economical to build e.g. a tank on a tower.