But technically, the processing gain comes from the filtering. The decimation is simply done to eliminate redundant data the filter created in order to save on processing resources.
The IIR filter (an R-C) is what is usually done to the input before sampling to prevent aliasing of the higher frequencies into the passband after the sampling and decimation.
I used to have this pretty well implanted in my brain, but I seem to have forgotten some of the whys and details.
Is he trying to preserve the data oversampling after the filter for some reason?