This seems incredible -- that products are chosen on such an irrelevant-in-the-long-term basis
I don't think there is anything too incredible in that. If you want to throw together an idea quickly, get it out there and test response then use whatever technology gets the job done quickest. You can always change later.
Why waste huge amounts of time setting up a technically perfect database for a product it turns out no-one wants?
The difference between getting competent with Postgresql versus MySQL was just a few hours. In the scale of a project such a difference dissolves into complete irrelevance, yet it was enough to sway many to use MySQL when it was severely deficient comparatively (though with its adoption it saw love that brought it up to if not beyond parity).
The same is true with many technologies and approaches. Projects that consume thousands or tens of thousands of hours, with a toolset chosen because it represented an outset savings of single-digit hours.
Changing in the future is seldom as easy as it seems in those early days.
The difference between getting competent with Postgresql versus MySQL was just a few hours.
Yes and no. There are a vastly more hosting options that provide MySQL vs Postgres. So it's not an issue of getting competent with the DB system, it's an issue of using an existing LAMP stack or having to roll your own.
I don't think there is anything too incredible in that. If you want to throw together an idea quickly, get it out there and test response then use whatever technology gets the job done quickest. You can always change later.
Why waste huge amounts of time setting up a technically perfect database for a product it turns out no-one wants?