What do you think you are going to improve by using
more channels?
Since only one at a time is selected, you will not be
improving on-resistance. The inputs do not act in
parallel for "on" conduction.
They do however act in parallel for shunt and series
leakage, capacitive loading. This is hardly ever beneficial.
Maybe paralleling could bring more ESD clamps into
play, if you are up against harsh input overstress
threats. But this seems like it ought to be handled by
something more explicitly suited to the role.
Now maybe redundancy could be useful to mitigate
intrinsic front end offset and drift (take the median
value, discard outliers) if each input has its own front
end amp in the "IC". but if the front end scheme
is all switches and a common front end amp / buffer,
no help there - switches don't contribute to offset
unless something is way bent. Not knowing anything
useful about this "IC" there's no way to judge that.
I would just tie the unused inputs to a ground or a
guard signal (if very low current) and put the guards
interdigitated with the signals so that signal-signal
coupling is reduced.
I don't have an interpretation of the problem because
the problem has not really been stated in a way that
tells me what the desired outcome, and the known
input challenges, are.