You're also making assumptions here. You're assuming that the query planner will always chose the right plan unless a human tells it otherwise (therefore bad plans are the fault of the humans).

Assuming that in the author's example it's really the fault of the human the wrote the query with a hint in it, it's not an argument against allowing hints. Just because someone might misuse a tool isn't a reason for not creating it. It's the same argument that the RIAA/MPAA uses when they want to ban P2P technologies (i.e. It can be used to infringe copyright, therefore it should be illegal).

Most of the 'hints-vs-no-hints' arguments devolve into two camps:

1) People that want to have control over the query because they don't fully trust the planner (this mistrust could be logical or illogical -- i.e. people that experienced a need for hints, or people that just don't trust the planner even though it's never given them a bad plan)

2) People that put 100% trust in the planner and say that since the planner is deterministic, then it's a human's fault when it comes up with a bad plan. (i.e. They love The Planner. They trust The Planner. They are careful not to taunt or anger The Planner.)

Hints can be useful not only for frequent queries, but those one-off queries a DBA can fire against the database where he knows how the query should be done. I know hints can be misused, but there are times the computer must trust me.