Dear Ekaterina,

I entirely agree that Python is a much better (and free!) environment for scientific developments :-) However it will take years if not decades to get an equivalent set of toolboxes in pure Python as we do now in MATLAB... There is however some middle term hope if toolboxes move to the opensource Octave language, an equivalent to MATLAB but free. SPM's team seems to be working towards it, so we can hope this might give some momentum for all other scientific toolboxes once SPM reach a stable status (although it can take longer than just porting to Python, which is also in the works!).

About running on a cluster, it should be possible to do so also from the GUI without compiling, indeed CONN provides such an option when one click on "Done", there is a dropdown menu to select the machine to do the processing ("Local processing" being the default). However I have no idea how to configure that, but if you show CONN to your cluster managers, they might be able to help in the configuration.

Hope this helps, and thank you for reporting in details the extent of the variability, this is very useful (you should consider publishing this result!).

Best regards,
Stephen