There are also other issues with interferential filters. Most bandpass filters and I assume thin film polarizers (I couldn't find any real manufacturing information on Cokin filters, but Wikipedia seems to agree with what I believe) rely on interferences to do their jobs. The problem with interference is of course that the angle of incidence of the incoming light matters a lot (astronomical filters for example are often provided with a range of acceptable F/D ratios): using a wide angle lens wide open is likely to cause all kinds of fancy effects. Here is for example a shot taken with an objective interferential filter at 24mm 2.8
Using a filter further down the optical path where the light is more perpendicular to the sensor plane (for example in front of the sensor like in the EOS clip filter families) works better for astronomical applications, but it is a bit hard to turn a polarizer in that location...