I would expect similar... imaging subsystems with these requirements would seem to be a low-volume/high-price type of thing, and that's not really SX's bread and butter.
Yep. SX
could build out internal capability and get into (for instance) the optical/SAR imaging game. They could try to leverage their brand (high volume, low exquisite-ness, fast iterations), but...that's not really where the money is in imaging. Planet has made a decent go at low-res imaging with their gazillion cubesats, but the real money starts in the sub-meter realm, and even then it's not really been a boon, as evidence by the various players in that space that are all more or less scraping by. You have to go to high res stuff to find a real cash cow business case (like Digital Globe/Maxar), and then you're spending tens of millions per sat to get the exquisite performance in satellite control and resolution. For some bench racing, the Worldview Legion sats are somewhere in the $20-40M range, including a $10M (or more?) Raytheon telescope on them (that's got decades of design heritage in Black Space...), a couple of million dollar star trackers, and CMG's that can snap the thing around at 30 deg/second and stop on a dime with full compliance, while returning probably 20cm resolution. (They say "sub 30cm"...)
Yeah, I was thinking that the "intra constellation" application would be more along the lines of Starlink sats being used to allow sats within a constellation to "relay" data from one to another if their density wasn't sufficient to do it themselves.
Ah, yeah. If you're hooking your satellites into the starlink ISL network then presumably they don't otherwise have the ability to comm with each other. However, I'd still imagine that any comm from Sat A gets "internet-ized" and then sort of vanishes into the starlink traffic before being routed back out to Sat B. Presumably if A and B are physically close the Starlink Routing Overlord Machine would decide that the smartest way to get that packet moved is to just hop through starlink sats and then back out to B, but I could also imagine the packet also getting routed on the ground for some part of its abstract journey from A to B.
What novel applications can you have if you have half a dozen sats with sensors that can cooperate directly? I dunno.. maybe nothing you couldn't do with ground-stations, but avoiding that as a requirement may be advantageous...
It's certainly fair to contemplate applications through the lens of The Future; its certainly reasonable to assume there will be use cases that we aren't smart enough to think about yet.
Today, there are sciencey missions that have some coupling of multiple satellites...but also there's no money in science so that's not exactly a market. Also the necessary coupling is more things like ultra fidelity ranging/sensing rather than data transmission. (How close is that other satellite, what is the atmospheric composition between us, etc.)
One could imagine heavy processing on a satellite...servers, mining (Blockchain!!!), etc...but even then intra-constellation comms still feels like the relay flavor rather than the cooperation flavor. One application I can think of here is imaging data, and the problem being solved is the size of the pipe to the ground: To varying degrees, [some] imaging sats on orbit do some level of image processing prior to beaming to the ground. It's mostly throwing out bad data and the like, but some do a little more post-processing depending on the request. I could imagine a future where a significantly higher amount of data than is collected today might benefit from significantly higher degree of processing, where you really want to start being really cognizant of what you're trying to squeeze into the ground link pipe. Add on potential desire to have a fast timeline between shooting the target and resolving that data into an action and local or near local computing might be useful (like if the next door sat is some big processing machine). We're still just talking data between sats at that point, but potentially much better enabled by keeping activity to the 'neighborhood' level rather than some terrestrial mothership.
One could also imagine some kind of autonomous defense application--Blackjack is kind of like a baby step there, and tying into above, at one point DARPA was very interested in using the Telesat light speed platform for Blackjack. The SDA activity is similar, where there will be multiple flavor of satellites in the constellation with various objectives, and there's some degree of on-orbit processing and "sharing of findings" amongst the satellites without ground intervention. Obviously there's no final action contemplated in those autonomous systems (and hopefully there never is...), but certainly/hopefully building a system that brings only useful data to The Deciders is going to be more efficient and accurate.
Relative to your thread pulling, the irony in the advent of optical ISLs (SDA now calls them OCT because--I assume--they needed a new acronym for a thing that already exists) is that the significant throughput and constant access to all the constellation nodes they enable sort of minimizes the need for satellites themselves to have massive on board processing capability or the ability to have significant PTP comms with their neighbors. It really just does become a problem about the size of the ground link pipe.