We have quite a bit of experience with the "giant packet" problem,
observed on SunOS machines.  There are some interesting issues:

(1) The reported "source" of the giant packet is usually NOT the
    party to blame for the "giant" packet.

(2) The actual guilty party is usually some machine with a bad
    Ethernet interface, such that it will begin transmitting a
    packet before some other station has gotten off the wire.

It can be very hard to determe who is stepping on perfectly
valid packets to turn them into "giant" packets...

FYI,
Gordon