I think it was the need for manual vacuuming. If you were running a small web host at the time picking MySQL over Postgres meant a huge reduction in potential support tickets related to non-vacuumed databases causing customers to run out of space.
That certainly didn't help, but if your DB was large enough where vacuuming mattered, you probably found a cron solution. The other place I saw MySQL win was in single access benchmarks that were scattered across the Web. Postgres would win in concurrent scenarios pretty handily, but people were glued to those serial benchmarks.
