Did you happen to leave a large portion of it on the spool? My understanding is that if you are not using a great length of the cable that unwinding a portion off of the spool will prevent errors. YMMV.
I heard that too, but the engineer in me would really challenge that. Anyways, pulled almost all of the cable out - same problem. Shortened the cable (120 feet shorter) and left it all on the spool, NO problem. This is also a sample size of two as far as the fix goes, sample size of 3 spools as far as the errors go.
Everywhere that I read about the physics of Gigabit and cable length - 120 Meters is pushing the design (or by some standards, well beyond the design limits). My scenario validated the "theory" quite well. It is also hard for me to believe that the console or AR's are at fault since it was consistent across an AR2412 and two AR84's plus with just the shortened length, 0 errors.
Anyways, we are happy now and have run 6 large events with 0 errors.