Well, so far, it's "a learning experience" 🙂
Generating a new MAC via the VRID: doesn't work too well. Yes, I have a made-up MAC, and yes, it's unique, but the switch doesn't want to get tricked. I presume packets stays on L2 and don't want to get forced to L3, because the switch is just too smart.
Sumit's approach: I can't guarantee the presence of "200.1.1.2" on that network or on that switch, so I can't use it, and I can't use the VR's IP because I run into MAC issues again. Or at least I think that's why that failed.
I ended up using our 480s to act as an external router between virtual routers. So I have one network/vlan "vlan_test1" in "VR-Test1" (172.18.0.0/24) and a vlan "vlan_test2" in VR-Test2" (172.18.1.0/24), tag them both to the 480s.
On the 480 I have "VR-XCHANGE", with both vlans (test1 and test2)", ipforwarding enabled.
Since we're talking about two 8800s and two 480s (yay, VRRP!!!), I did accidentally assign the same VRID to the VR on the 8800s and the VR on the 480s. Bad bad bad idea, because all of a sudden they all thought they needed to be able to be redundant, but they saw different virtual IPs (Of course! One for the 480s, one for the 8800s) and freaked out. Changed the VRID on the 480s and voila! all is well!
Now I know what else VRIDs do, other than change the virtual MAC I think, Paul, you mentioned that in another thread about VRRPs - thanks, saved me today.
It would be oh-so-nice if I could've just assigned MACs to ports and used a patch-cable, but I understand the design history/decision.
Thanks for everyone's input - and if I ever find another way to skin that cat, I'll be back in this thread 🙂