I also found
this page.. Author did much better job, testing more brands.
Celestron 0.63 test result is consistent with Damien's specimen, but my result is different.
I also found that behavior in frame corners (APS.. FF will be even worse) is not much dependent on working distance...
It seems that the "working distance" value for FF-FR everybody is concern with only ensures f/number (which is pretty unimportant parameter actually) but not the flat field (it actually worsen it).
I think in the next step of wasting my time I will try a simple doublet from old binoculars (FL =~21 cm).