This is a VBT (Very Bad Thing). Unless the non-STP aware switches (or hubs) are tightly controlled they are likely to cause spanning-tree loops. The first thing you MUST do to prevent this is to turn on "spanning-tree portfast BPDUGuard default" if your switch's supports it and then ensure that the ports these switches are connected to are setup as access ports with portfast enabled. That way, if an unmanaged or non-STP aware switch or hub gets plugged into your network twice creating a loop, one or both of the links will go error-disabled after a brief delay, hopefully. It sounds like this may be your problem.If your switches do not support BPDUGuard (and your switches are all old, EOL models so I don't know off the top of my head if they do) then I would HIGHLY advise you against it unless you are willing to suffer repeated outages.
The above settings are considered MANDATORY by Cisco best practices and rightly so. In large enterprise networks unmanaged hubs/switches are hunted down and terminated with prejudice :-). We had a user install one of these recently and bring down a complete data-center even after precautions were taken to prevent just that.
Nathan Spitzer
Sr. Network Communications Analyst
Lockheed Martin
