A flat multi element phased array is still better if you aim the flat part in the general direction where you want the most gain. This is the why the flat phased array radar in the nose radome of a modern air superiority fighter jet is mounted on a motorized steerable platform.
Starlink dishes do have a steerable mount. It just usually doesn't move in operation. It might be fast enough for on-the-road use, with fine tracking done by the phased array combined with coarse tracking with the motors. That said, since they don't move in operation anyways with moving satellites, I think they should be fine on a moving platform with just beamforming.
There are a lot of Starlink satellites, and they move in a very predictable manner. I wouldn't be surprised if, for stationary use, it actually just picks a very small cone of the sky and only uses satellites from the same orbital plane moving across that cone along a circle segment over the course of multiple seconds.
Moving the dish would then only be for optimal array alignment with the "target cone", not for tracking individual satellites.
This seems very different from the movement of e.g. a boat in high seas, which can roll or pitch by up to 30 degrees over the course of very few seconds.
the motors in current model starlink terminals are not the robust type that would be used for constant-motion tracking 24x7x365 (as you see with an o3b terminal for MEO), they're motors that are intended to be used rarely in a fixed mount application where it aligns the panel at a certain heading one or two times a day, maximum, and stays that way.
