In our case, it was because the older versions of Postgres would quickly lose the ability to use indices on oft-changed tables. This, of course, meant the query engine would be reduced to sequential searches. The only way you could fix it was by running VACUUM ANALYZE which could take forever and a day on large databases. Then, it would run great until it "got dirty" from insertions and deletions and you had to do it again.

Contrasting that with "Works with no performance degradation for years on end" made the decision to use MySQL a lot easier. Wasn't all sunshine and roses though because, at the time, MySQL was nowhere near Postgres in terms of features.

NOTE: This was sixteen years ago. I understand it is leagues better now but that experience really soured me on Postgres.

// @kdfrawg