We'd had various high tier and even "Crit Sit" cases with IBM for over a year, and feel we're now getting to the bottom of all the issues we've had with IBM blade SAN connectivity.
Our first problem was a previously unknown incompatibility between the Brocade 8gb Bladecenter switches and our quite old 2gb McData ED140 core switch. This problem caused random blades at random times to throw tons of fiber port errors (CRC's, encoding errors, etc). We pulled our hair out for years but eventually upgraded our core to newer Brocade 8gb switches. Which eliminated the first problem but exposed a new problem: Blades in certain slots just performed poorly, and would occasionally spit CRC's, ENC errors, discards, etc. As it turns out, near the end of 2012 IBM had diagnosed that half of the 14 blade slots in all BladeCenter chassis do not have optimal connectivity length between the blade and the switch. Some are too long, some are too short. Slots 3 & 4, and slots 10-14 as I recall. Our experience has been that slots 4, 12, 13, 14 have been the most commonly bad. IBM has a new midplane available that is supposed to mitigate this issue. As fun as replacing midplanes in many bladechassis' sounds...
Another good discussion with a group leading to the same conclusion: