[[Mod. note -- It seems to me that a plasma effect would be (strongly)
frequency-dependent. Astronomical redshifts are independent of frequency,
so I don't see how a plasma effect could produce them. -- jt]]

As indicated in my opening post, one should note take the comparison
of the redshift with a dispersion effect too literally. I mentioned
this merely in order to show that a suitable mechanism could deform
the wave packets such as to lead to the redshift and still conform to
the constancy of the speed of light.
The usual dispersion theory can in fact not be applied in case of the
intergalactic plasma because not only is the average distance of the
charges much larger than the wavelength of light but even much larger
than the coherence length of the wavetrains (the coherence length of
light emitted from the collisional plasma in the photosphere of stars
is around 10^-2 cm whereas the intergalactic charges should have a
typical distance of 100 cm or more). For most of the time there isn't
therefore even a single charge within a 'photon' length. However, the
latter is then still in the electric field of the charges which could
lead to the redshift (it may actually be the field gradient which
causes this).
Again, this is would be a new phenomenon which can not be derived from
the usual dispersion theory as it is outside its scope. Even without a
quantitative theory for this, it could be tested observationally
however as for sufficiently long wavelengths and/or coherence lengthts
of the light, the redshift effect should be reduced and eventually
disappear altogether (in my opinion the Cosmic Microwave Background
radiation might be an indicator of such a threshold). One might also
be able to demonstrate the effect in the lab by observing the
propagation of light in a suitable static electric field.