There's only one PostgreSQL incident story, retold countless times. There's a bad query plan, we know what the right plan is, the tools to detect the problem are poor, and PostgreSQL won't let us force a fixed plan, so we have to do some unholy hack to nudge it towards the right plan. They even relied upon another pretty substandard aspect of PostgreSQL: the fact that CTEs are optimization barriers. For god's sake, it's 2022.

On SQL Server, we see it building the bitmap in the graphical query plan basically right away. We can see it in Query Store, and we can see it in sp_whoisactive. The actual query plan being used in production is listed. Then, a couple minutes later, you've deployed a query hint and the bitmap is gone. This is not a hypothetical scenario I am imagining; I have done exactly this, for exactly this problem with the query plan trying to build a bitmap on a huge table. This is a minor blip for SQL Server users and not a disaster that turns into a blog post.

At what point does PostgreSQL change their minds about officially supporting query hints in the base system? This isn't the first time I've read basically this exact blog post and then written this exact comment on HN.