Firecracker:
Tough question. Has this particular switch exhibited any problems in the past. Is it new (factory default settings, no vlans)? I assume that you are using cat5 to cascade the switches, or are you using some type of stack cable to integrate the switch fabric? Managed switches use an algorithim called spanning tree to protect against loops and broadcast storms. It doesn't take more than one bad actor (NIC going bad, wonky wire) to create a problem. Have you tried just moving the problem switch to another table, just to see if the problem persists with a new position in the cascade. Sorry it's been a long time since I've actually had to troubleshoot any layer two issues. Wish I could be more help.