I'm pretty sure this has to do with subdivision of reflected surfaces which are outside the camera frustum.
If surfaces are outside the camera frustum they are subdivided less as an optimization, however, these less subdivided surfaces are being reflected inside the camera frustum and thus look facetted.
What we need is to have TG2 subdivide surfaces outside the camera frustum so that they appear correctly in the render.
In the rendersettings panel go to the advanced tab and set the "ray detail region" mode to "detail in camera".
Then set the ray detail region to 1.
This setting will extend the region of calculated subdivisions of surfaces to 9x the size of the camera frustum.
(1 = 100%, so 100% extra on left, right, bottom, top and left/right bottom/top = 9x bigger with original camera frustum centred.
0.5 = 50% extra added frustum.)