These switches won't take the "ip load-sharing per-packet" command on the FE interfaces. I did some tests with another 3550 and 3560 and even though it took the commands, all of the traffic still went over one FE interfact.
I tried different methods of "ip cef load-sharing algorithm", too.
I tried making the two L3 switches OSPF neightbors and even though max-paths was set to 3 and there were 3 equal cost paths to the tunnel destinations on the "ASA"s, all of the traffic flowed over on FE interface between the switches.
The only way I can test this is to open a couple ping windows on a laptop connected to one "ASA", which is really just a router. The two routers have a GRE tunnel through the switches, so the only traffic the switch sees is GRE traffic to/from the same ip/mac. I confirmed this was the only thing going out the interface towards the switch by using "show ip cache flow".
Three trunking ports didn't do any good, either. I assume two of them would have just gone into blocking, anyhow.
Thanks for any help. :)
-John