[c-nsp] Odd multicast behavior from an ME3400

[ML]

Using an IneoQuest cricket we've been trying to find out why multicast 
video streams are breaking up.

Using an ME3400 as an access device these are our symptoms:

3 x MPEG4 HD streams (8-10MBps each) come through fine.
Add one more stream and the Cricket says we've got problems.
On aggregate this about 40Mbps/3500pps with four streams.

However with *10* standard def MPEG2 streams (~84Mbps/7500pps).
Everything looks good according to the Cricket.

If the access device is a 3560 we can pull twice as many MPEG4-HD 
streams without issue.


We've already verified the content is good from the source (pulling 
twice as many stream as a customer would ever pull)
now it's down to the ME3400s in the access layer.

When looking at the interface counters there is not a single error of 
any kind. "show buffers" shows no changes in buffer misses
during and after the point where video breaks down on the MPEG4-HD streams.

When the Cricket's video monitor point is a gigabit port (via a GLC-T, 
albeit the Cricket monitor port is 100Mb) we can pull four streams with 
problems but the error rate is reduced.

We are using the IPBASE image. Tried several versions 12.2(25)SEG1, 
12.2(44)SE no difference.

The multicast config is basic at the access layer.  Just default config 
for IGMP snooping on the multicast VLAN with immediate-leave.

I used the Bug Toolkit but nothing stood out to me as an open/fixed bug 
with our symptoms.

Is there a troubleshooting step I'm missing here?

Thanks