I would suspect a mechanism in the ortho-mode
transducer (OMT) that exhibits a combination of
a slightly elevated VSWR and some mode/polarization
conversion. In addition, the currents flowing on
the dish surface will contribute to some of the
increased noise level on the opposite polarization,
as will backscatter from the sub-reflector (I imagine
there is one).
If you analyze the expected levels
on the degraded channel, from knowing a set of parameters
such as return loss, CP Isolation and the channel
bandpass filter characteristics, in terms of dBc,
you may be able to see what deterioration in any
of those would lead to the increase in noise floor
you are seeing. One of the elements may be of particularly
high sensitivity, so a very small departure from
ideal will change the unwanted noise significantly.
Only then can you address the issue of what to do
At bottom, life is all
Sucking in and blowing out.