What I mean by Router Teaming is really just that I'd like to "daisychain" two routers, so that one becomes the "Parent" and one becomes the "Child", and they work in tandem and present the same Configuration (which applies to both.) I could then, say, wire two 2011s together and get two wlan interfaces and 18 ports (+ 2x SFP)

Hypothetically you could run a long chain of "Child" routers, but I see this more as a use case for small-scale upgrades or capability expansion.

There would obviously be limitations on routing speed between ether1-master and ether11-master (or parent-ether1-master and child-ether1-master)

Working together is very wide range of possibilities. Can you please exactly specify what the feature should do in details? Why you cannot make such configuration with actual set of features?

Isn't too early to request such feature without deep knowledge of the features you already can use?

I'm not talking about:
Interface teaming / bonding,
Managing multiple APs, or
Connecting to other routers over SSH

I'm talking about merging two (or potentially more) physical routers into one logical router, so that all the interfaces appear to be part of one device. The LCD panel on either device, for example, would show stats for all interfaces. This could be largely a case of doing some fanciness in Winbox/Webfig.

Ok. If you think so...

Hey, it's just an idea! I don't know the architecture of RouterOS, but I could imagine that if the API also drives Webfig and Winbox (which would make sense) then you could quite easily build an aggregation façade to pick up a composite view of the "clustered" environment from any node on a call-by-call basis. Then the tricky bit is configuring a new node and making it part of the cluster.

Since there's only one disk per router, there could be some complications in running scripts. If you wanted to leverage the total processing power by distributing the load, you'd need to sync the disks across all nodes. Alternatively you could potentially create a virtual array of disks. Any more than that and it starts to get a bit mental.

This sounds like "stacking" switches.
Switches that support stacking have a pretty high-bandwidth stacking interface so that your total available fabric bandwidth is sufficient to make a "single switch" out of the various shelves.

I would find it much more useful for state mirroring to be implemented - i.e. N+1 chassis failover redundancy. This way, if box 1 dies, box 2 takes over immediately with all of the connection state tracking, dynamic configuration entries, etc - all in place for a completely seamless failover.

Yeah, what you're talking about is clustering. Treating two physical routers as a single object that has built-in failover/redundancy and/or load-balancing capabilities. Almost every major routing product out there supports this...Cisco, CheckPoint, PaloAlto, etc. It definitely simplifies configuration, as you only have to set it up once, versus configuring each individual router like with VRRP. This kind of setup would be great for core networking, where you're dealing with first-hop from critical network services that require full uptime/availability, or great for transit nodes in an ISP WAN. More traditional first-hop redundancy technologies aren't fast enough to use in your core. You could make it work for transit nodes in a WAN, but that would depend more on how many alternate paths exist in your WAN mesh.

I'd personally love clustering just for the simplified administration.

Yep, so "Stacking" and/or "Clustering" is really what I meant.

You'd need the high-speed interface where you want to, for example, set a master port for an interface on a different router, or include a port (or master port) in a bridge or some other configuration that would necessitate relatively-high-speed traversal.

Perhaps there's a "simple stacking" that could use Gigabit ports, and a more complex mode that would require 10Gb SFP modules?

For "simple mode," the obvious mechanism would stacking, such that (for example) parent-ether2 connects to child1-ether1, and child1-ether2 connects to child2-ether1. I'm not sure there's much latitude in supporting a "star" topology where all child routers are connected to a Gigabit port on the parent.

When using 10Gb SFP modules, would you mandate the use of two modules to "stack" as above, or would would it be possible to figure out a "token ring" type arrangement? (The latter would mandate the use of single fibres terminated with single LC connectors.)

One of the best clustering setups I've seen is what CheckPoint does with their security gateways. Their management software allows you to deploy your firewall rules to the cluster (so it installs on all cluster members simultaneously), as well as manage individual member configs that get pushed at the same time as the rule bases. Most people typically reserve one physical port per cluster member to act as the sync port. If you only have two members, you just run the cable directly between the hosts, and all state synchronization and cluster member health is monitored over that link. If you have more than 2 members, you run that physical port to a group of VLAN'd ports on a switch so all cluster members can communicate with each other.

It would definitely be nice to have some sort of clustering capability/unified management to VRRP, along with state synchronization and automated failover.

Why not use OSPF for that? That protocol is designed to have a fail over when one router failes, the routes are automatically redefined.

Or maybe you need a physical failover? Well, if the router fails (what is failing in your perception? Like power adapter failure, logical "lock up" etc...), then there is no help for the people behind it.

Or are these routers on the same switch?