I have RNASeq count data from 42 different conditions, 5 repeats per condition for a total of 210 samples, and approximately 37,000 genes. DESeq freezes up at the estimateDispersions step unless I slice the data down to ~50 samples or so. Is it the case that the calculation time scales up in the extreme, or is it something about one or more of my samples causing the freeze? I have tried parallelising the step with 120 cores, but to no avail, and I feel like it's perhaps a crash rather than just a long calculation because I can't cancel it (though that's often the case in R so who knows).

Any advice appreciated.

It's not "freezing", but just going slowly because the design matrix is very large (both number of samples and number of parameters). I'd recommend to use limma-voom.