I added a transform input set to worldspace, but it didn't seem to make any difference. At least not here.
TG file attached. There are three cameras, with one near the world origin and the others near x = 20,000 and x = 60,000. Tested various water surfaces including a world-size sphere, a single lake at 0, 0, 0, and individual lakes and planes at x = 0, x = 20,000, and x = 60,000. You can turn them on and off to test the various combinations.
The attached image shows the results from the three cameras using the sphere and individual lake objects. Using a single lake at 0, 0, 0 for all three cameras gives the same result as the sphere. Using individual planes at the three locations gives the same result as the individual lakes.
This is all consistent with the other types of distortion we typically see with positions and displacements far away from the world origin.
The lesson I take from this is to use a sphere for water surfaces, at least for large scenes or for those distant from the origin. (There are reasons why I need to use a plane for my current scene.)
Didn't test different y values yet for the surfaces. May get to that later.