I'd expect the steering to operate well beyond tens of kHz. It should be able to update aiming at least within a few transmission time slices, ideally after every time it receives (and that can include anyone else near enough for the satellite's steered beam to cover it) it can fine tune the aiming. Keep in mind it will be keeping in touch with two or more satellites most of the time to perform seamless handoff as the satellites fly by, so it already must be able to quickly steer between satellites which might not even be on the same orbital plane but completely perpendicular. All this steering is digital and not mechanical, so the only limitation is how fast they want the steering algorithm to update, and thus how much CPU power they're willing to give it. It's not a super demanding task, relative to the amount of data that can flow through the link.
And even if you have trouble with a dinghy on the open ocean, there's no need to nit pick between "boat" and "ship". Any floating vessel big enough to have a use for starlink (versus just having a Iridium sat phone for emergency rescue situations) should be fine. These are not your grandfather's satellite dishes, so to speak. There is nothing to move, nothing to aim, just simply digitally controlled beam steering. You don't even have to be particularly careful about what angle you attach the terminal at, it doesn't need to be on a flat surface, just have a reasonable view of the sky, and it will handle the rest.
The 10s of kHz was limited by the off the shelf $5 hobby gyro/accelerometer I looked up.