One choice puts a lot of risk on the users of software. One choice puts way less. You're constantly encouraging putting more risk on users and exploits into arbitrary attackers' hands with the argument that one or more attackers might already be a risk. That makes no sense. The only people with a straight-up benefit from this are malware writers who don't want to work very hard converting the patch into an exploit.
Researchers should delay the publishing of exploits at least a week so people can test and deploy the patch. I keep mentioning testing because patches sometimes break stuff themselves. Doing otherwise just saves attackers work.
Instead, you want to acquire immediate vulnerability to a large group of hackers to prevent a patch from maybe making you vulnerable to a smaller number that either has an existing exploit or RE'd the non-working patch to devise a new one. Makes no sense lol... Better to get the patch, do basic tests to ensure it doesn't break functionality, install it, monitor the system/network, and then verify the security later when the sploit is released.