2. Failovers of any database are not instant, but they are indeed quick! So let’s not claim that you can do an upgrade with zero downtime.
3. In-place upgrades are extremely fast and you can test the speed using a physical replica before hand, usually it’s a couple of seconds though the docs say minutes.
4. MySQLs major version being in the minor position is exactly the kind of “you should be sure you know what you’re doing but we won’t make it obvious” territory that I really despise.
It was nevertheless a pretty epic journey of dialectic discourse plunging _deep_ into the esoteric and nuanced realm of expert-level technical minutiae. A mostly intellectual journey, albeit distinctly punctuated by an undertone of emotional angst that steadily progressed in its growing intensity in a manner proportional to the magnitude of your collective disagreement… epic indeed.
I'm responding to you because you're repeatedly posting factually incorrect items, for years. For example you and I have directly discussed the "MySQL doesn't use SemVer" thing before on HN, and yet here you are again in this thread, claiming 5.6 to 5.7 should be a "minor" upgrade.
Anyway, to the topic at hand, as others have also mentioned in this thread: historically the difficulty with Postgres upgrades has been the lack of cross-version replication, due to Postgres WAL replication being a low-level physical replication system. This made it difficult to perform an upgrade while keeping your site fully online. Perhaps the newer logical replication support makes this easier these days. I hope to learn more about it someday. If you can share your process for upgrading a Postgres cluster while keeping it online, that would be helpful and informative.