I think people are overly quick to dismiss the effectiveness of IDS. Yes, rules often match symptoms/exploit instances rather than actual vulnerability semantics (though this has improved. But like a bike lock, even imperfect IDS gives you protection against opportunistic attacks that are pervasive for any system connected to the Internet.

The fact that IDS can respond without needing to validate a new software release means it can also very often outpace remediation through software updates.