For people who have budgeted, procured and installed 'serious' two way geostationary stuff in the past (one example of which would be a 2.4m two port linear compact cassegrain antenna, NJR PLL LNB, a 40W BUC, a Comtech CDM760 modem and a 1U sized Cisco router), for 1:1 SCPC dedicated transponder capacity based services, starlink is atonishingly fast.
I could pull out a check book and spend $45,000 on buying terminal hardware and $30,000 a month in transponder space and not be able to achieve the speeds that starlink can do right now. Even if starlink was only 20 Mbps down and 4 Mbps up, go price what 20x4 service will cost by traditional geostationary right now (hint: start looking at $1200 per Mbps per month and multiply by N number of Mbps).
It is indeed a good theory that I'm seeing unreasonably higher than normal speeds right now and better latency, jitter and packet loss because I'm in a similar situation to being the first customer on a new WISP PtMP AP sector. But I also have a great deal more confidence that spacex's continued paces of launches and satellite deployment will keep up with providing at least a 100 Mbps down x 15 Mbps up service. I do not think that they will let it degrade into a contended-service-hell where customers see a very poor end user experience.
My perspective on starlink is also informed by knowing the price right now for Inmarsat and Iridium based offshore and aviation data services (sub 2 Mbps) and the $ per megabyte costs. There's already starlink aviation terminals in beta, and terminals for maritime and offshore use. It'll be a game changer there. The market for a globe-covering LEO high throughput satellite network is much larger than just the US48 state consumer residential internet/small business last mile internet market.