We are having alot of TCP performance problems, and I believe they are attributable to our HSRP implementation.
Imagine 3 routers on a Token Ring segment. RouterA has static routes to HSRP virtual address shared by RouterB and RouterC. RouterB and RouterC have WAN links to the destination addresses. (We don't use a routing protocol here because RouterB & RouterC are service provider managed devices and frankly we don't trust them enough). When the static routes point to the virtual address, we have significant TCP performance problems for all WAN destinations. When we change the static routes to use the real IP address of either RouterB or RouterC, performance is perfect.
Although I have been unable to run HSRP debugs on the managed provider RoutersB &C, the config looks standard and a sniffer packet capture verifies that the HSRP multicast packets are being send and received and the states are valid and stable.
The only other potential issue I can think of is the use of Source-Route-Bridging on Routers B&C Token Ring interfaces (used for SNA) that have the HSRP config.
The following links provided some insight, really don't offer a solution:
Thanks, Tom