After a cursory glance at Cloudflare D1's description and service limits, I was left with the impression that the usecases are not that of a generic RDBMS. I mean, the limit for each database is 10GB, and apparently a D1 database resides in a single location. Of course this means >200ms to even hit the server when requests don't come from the same data center.
Perhaps this is one of those things where a underperforming implementation is released to minimize time-to-market, and then they work on performance. However, for a distributed service this suggests to me that this solution only works as an ad-hoc storage of complex relational data for high-read, low-write scenarios, where you have Cloudflare Workers query D1 but either aggressively cache the response or store queries in Cloudflare KV to avoid the roundtrip cost.