Scenario...
I want to use a streaming client/server video device across an EOIP tunnel for my two residential places. Said system is very broadcast and subnet sensitive. As far as I know, only connections at the level of EOIP have been successful.

I don't really don't want to filter all internet traffic from either site through it's peer. This takes higher priority than network design purity. While the link is good, it's not good enough to co-exist with remote desktops sessions to internet hosts.

So the 'simple' solution feels like EOIP with split DHCP servers serving the exact same config with split pools of 10-127, and 128-254. (and sadly, manually dual-maintained static IPs for the couple clients that need it).

I'm not thrilled about this approach. Am I missing a more obvious and less brittle solution space that would meet the requirements of 'shared for broadcast and Lan, but independent for off-subnet routed connections'?

Not an answer but that DHCP setup is not necessary. The ISC DHCP server has a primary/backup failover capability and it works fine. Just configure them both the same way and they communicate pool information to each other. The docs have some pretty simple setup information.

I kinda wonder why EOIP is so wrong if it's available to be used. I agree it's probably not right to use for most stuff, but if you really need a L2 segment between 2 separate locations, it's a way to do it.

Not an answer but that DHCP setup is not necessary. The ISC DHCP server has a primary/backup failover capability and it works fine. Just configure them both the same way and they communicate pool information to each other. The docs have some pretty simple setup information.

It's not the dhcp server connectivity/sync that would drive this decision. It's the ability to have each physical location use the local router as its primary gateway. Maybe there's another/better way to do this?

If you have to have the same layer 2 domain across both sites, then a split DHCP implementation is the best way to go. It ensures survivability for each site if the tunnel between the sites goes down. You should be able to achieve this with GRE, but if you already have EOIP working, I'd just stick with what you have. If it ain't broke...