Chris, we are an metro carrier running on a layer 2 core of 8900's and this has been a sporadic issue for years. Best work around we have found is disable IGMP snooping on your layer 2 only switches as much as possible. If you are not running video that needs to be pruned then just kill it 100 %...
Fix/workaround to help you find which switch you may be having issues with ( theory is a mcast entry get's corrupted in hardware and then not forwarded) ... You need to make sure that your two routers can at least ping their outer ip interfaces... If not then you have other issues.
One switch at a time when you have OSPF or router agency issues...
clear igmp snooping
clear fdb
clear ipmc fdb
check to see if your routers re-gain their agencies after each switch you clear in the path till you find the one that was at fault... GTAC will tell you which code you will need to be running for the switch and setup you have. 15. had some issues for sure.
We have found that this seems to be a very random thing and is usually triggered after a topology event where you have an EAPS failover. Seems to happen when we have port in the rings that flaps multiple times in a short time period.
Good luck, We have never found this to be a lack of resources so there are always open buckets in the memory for more entries. By default unless you have an ACL in place to block 224.0.0.0/24 all modern layer 2 switches should always forward the mcast traffic from router mcast ip's period. CPU only moves it to hardware first time... so good luck on your efforts... I will be tracking this one closely too for new info or ideas.