You and I think very differently about this. I don't own a SmartThings thermostat; I have one with an open API that I integrated with SmartThings. I don't have a SmartThings remote control; I have iTachs with an open API that I integrated with SmartThings. I don't have a SmartThings computer... well, you get the idea.
That's not really any different from the way I view the system as a whole. Let' see why:
I don't see lighting any differently. It is (and I prefer it that way) another switch on the network.
Here's where there's some difference between us. Your (and my) thermostats and similar devices are interfaced with via proprietary API abstraction layers because those devices don't support any standard, open protocols for directly exposing their functionality. While I also view lighting as "just another switch on the network" I realize that the bulbs themselves ARE network elements that implement and are accessible via an open API and set of protocols (Zigbee), and that operate as "End Devices" (and perhaps also as "Repeaters"/"Routers") subordinate to a "Coordinator" (the role filled by the bridges/hubs). Since the ST Hub includes that Zigbee Coordinator function, as well as all of the automation capabilities of the Hue Bridge (and more) the Bridge is a superfluous additional layer. If the protocol(s) between the Hue Bridge and the lights was/were proprietary, required some special software extension be added to the ST environment or was/were subject to change by Phillips on their own whim then using the Bridge as an abstraction layer, so as to insulate your architecture from suddenly introduced incompatibilities would make sense. But Zigbee (and Z-Wave, for that matter) is an open set of network protocols, just like 802.11, and having the ST Hub fulfill the roll of Zigbee Coordinator not only eliminates a pointless (in my view) additional layer (and set of communications hops, and one additional device to malfunction, etc), it also strips away the limitation of being able to use only those device features that the Hue API chooses to expose, and in only the way(s) that it chooses to allow you to use them.
In short, my Z-Wave and Zigbee mesh networks ARE a type of subsystem that ST is interfacing with. Adding the Hue Bridge as yet another abstraction layer to that architecture just doesn't buy my anything that I deem to be of any value.
My ST Hub also interacts directly with other Z-Wave & Zigbee devices as well (motion sensors, switches, etc.) Would there be any point in similarly inserting some other company's Controller device between them and the ST Hub? I don't see one.
While my Hues are integrated with SmartThings, Alexa talks to them directly; as does my scene controller (a Mac Mini). SmartThings can talk to them as well, but so does Tasker and iRule.
Ditto all of those things and my ST Hub.
Putting all of your eggs in a SmartThings basket (or any kind of singular controller)...
But that's not at all what I'm doing. It is not a SmartThings architecture. SmartThings is simply filling one role (controller/coordinator) in a HA ecosystem that consists of devices that implement various interface protocols, some open...some proprietary. Having ST interface with them as directly as practical means that I have fewer places that I need to go, and steps to go through, to add new devices to that ecosystem. Since the ecosystem is based on open standards I can use a different controller at a future date if I need to, or add more to it if there's some compelling reason to.
makes for fewer options (why open the painfully slow SmartThings app when I can just tap a shortcut on my phone's home screen)
I almost never use my ST phone app to control anything. I have automation rules, Echo, etc for that. The app is used for making configuration/rule changes. The speed of starting the app is a complete non-issue for me.
and a complete teardown/rebuild should something go unsupported.
I'm not sure how having a Hue Bridge makes any difference there.
I myself have learned over the last several decades that one protocol/device is the wrong way to go, and that integrating individual, open API based systems gives you a much more versatile and future proof design.
I've been a software engineer for ~30 years, and I learned that lesson decades ago myself...which is why I'm most definitely NOT building my HA on "one protocol/device", and am also "integrating individual, open API based systems". The only difference here is that I recognize my Zigbee mesh network and the lights that are a part of it as an open API based system.
One of SmartThings biggest advantages over its competitors is being able to easily integrate various subsystems; it be a shame to ignore it.
I agree. That's why I'm not ignoring it.
The added cost is a drop in the bucket in the grand scheme of things
Yes, the cost of the Hue Bridge is a very small part of the overall cost of a HA system. But it's a cost that doesn't buy me anything I want/need, and as such represents a bad value.
and as far as unnecessary, is every drop on your net connected to the same switch? Every WiFi node connected to the same AP? Of course not; every well designed network has branches. And while I suppose you could deem them as unnecessary, they have purpose.
Are you talking about a home TCP/IP network, or one set up to handle a substantial business? I have one active WiFi router in my house (with a spare in the closet in case of a malfunction) that handles all internet traffic, and several devices that communicate directly with one another via WiFi, Bluetooth, Zigbee and Zwave. I'm automating a home as a fun hobby, not building the next Space Shuttle.