1) The short lifespan is intentional. There is no reason to keep satellites beyond a few years because older satellites quickly become obsolete. The first generation Starlink sats are now 4 generations old, and are practically ancient at this point. SpaceX wants older satellites to fail, so they can be replaced by updated ones.
2) SpaceX could easily increase the lifespan of Starlink satellites to 10-15+ years, but they don't want to, because of (1), and because they are waiting for Starship to launch larger satellites at 10x lower cost.
3) Despite continually replacing satellites every ~5 years, Starlink is operating at a very profitable 40%+ profit margin.
4) Starlink is expanding to direct-to-cell 5G at speeds up to 150Mbps, which vastly expands the TAM from just broadband usage to all cellular data usage. Rural areas may never get a terrestrial cellular buildout, since it will be cheaper for companies to partner with SpaceX for coverage than to build their own towers. A single cellular tower can cost upwards of >$250k.
One thing that isn't going to happen is to magically extend the life of low Earth orbit satellites. You can have low latency, which you need for a telephone network, or you can have long lived satellites. But not both. You also need propellant to avoid collisions.
$250k is plausible for a high capacity cell site for an urban area. A small site with wireless backhaul can be $10k, or even less in some cases. If I am to believe the optimistic numbers, a Starlink bird costs about $500K and putting it into orbit cost of another $500K per satellite. But that means be believing that SpaceX can launch a used F9 for just $15 million, when the retail price of an F9 is $75 million.
EBITDA is the wrong yardstick when you're talking about space rockets having to deliver your infrastructure.
1) Why does Starship need to launch a 100-ton "proven" payload? Starship has already demonstrated it can launch dozens of Starlink v3 satellites into space.
2) Starship already has the heaviest payload and lowest cost-to-orbit of any launch vehicle ever, at $400/kg to orbit in its current fully expendable configuration. Why does it need to "be reusable without lots of expensive refurbishment" to show it's a success?
Even $400/kg to orbit is largely exaggerated; it assumes Starships are hand built and not mass manufactured, and also assumes no booster reuse, which was already demonstrated.
3) What do you define as "low latency, which you need for a telephone network"? The conference app Zoom requires below 150 ms latency, and recommends below 50ms latency. Starlink latency at its current altitude (~500km) typically ranges between 25 ms and 45 ms. Future satellites will orbit at ~330km, which will reduce latency further.
4) Why are "low latency" and "long lived satellites" mutually exclusive? My understanding is this is a adding enough propellant to a satellite, will make it long lived.
5) How can I get a small site with wireless backhaul for $10k? The lowest costs I see online are $20-30k.
6) You do realize the cheapest cellular backhaul sites at that price point are powered by Starlink? I.e. it's cellular equipment attached to a starlink terminal.