it was various kernel states. The guys matched the kernel configs /etc/system. /kernel/drv etc to match the machines that didnt have the problem. Didnt fix it. So they tried doing a core of a running kernel to use MBD to poke around. I said do a modinfo it will show what resources are being used and a diff will point out exactly which module is having the problem to narrow the search. Thats when they found the good boxes had a count of 220+ and the problem ones had a module count of 120 or so. That goes back to the system eng who said the wrong jump image must have been used. We use company modified kernels for various reasons.
thanx,
bill