[BioC] multi-experiment multi-chip summarization

[Anna Lobley]

I am trying to integrate several diverse affy datasets
from the GEO database all from the same organism
same chip design (~500 samples).

I'm interested in obtaining absolute expression values
and have created a robust target distribution for quantile normalisation
over all samples (carried out externally from R).
 
Due to cpu memory requirements I have only
been able to run median polish for probe summarization on the dataset
within single GEO experiments rather than across all of the quantile
normalised data in one go.

My question is fairly open ended and as follows:
I'm concerned that this methodology will over-emphasize
variance between experiments that could be avoided using
median polish over the entire dataset. Is there a more
appropriate way of carrying out summarization on this large
dataset?

thanks in advance for opinions/help