Ok, we finished the PRP check and reported the result some time ago. One of the cards had no error, and no reload. The other had 7 errors and about 20 reloads. I didn't understand exactly what a "reload" means, and I didn't bother. Because one card was faster than the other (better cooling on the bottom of the rack) we switched the cards periodically, to end the test in (about) the same time, therefore we ended up with one result file saying "1 error" and the other saying "6 errors", but all the errors were "nurtured" by the same card. Somehow, at the end, they both got the same residue, which was OK for us, and we reported the result to the server. However, to out great disappointment, the DC can not be reported, because the owl doesn't have a shift, all the tests are with shift zero. Which is bullshit, if there is any error in the code, both our tests operated with the same data, so both of the tests could be wrong, albeit they got the same residue.

Up to now, we assumed that this issue is only for LL, but the PRP is "properly" implemented. False hopes... We will revert back to "forzes" and mfaktc, putting the "sevens" in the drawers for now, till the owl will open its eyes and see in the dark as it is supposed to do. We don't like to walk around blind.