First of all, good opinion questions are good, no question about it. The canonical resource is Good Subjective, Bad Subjective. The issue here is really poll questions. I'd love to be able to create this kind of content, but I don't think our site is currently a good way to do it. There are two main directions we could consider exploring.
Accepting some poll questions
We could decide what kinds of poll questions are okay, as you've suggested. We have to do this in a way that's clear and easy to explain. Anything that'll cause debate on most poll questions is a non-starter. Anything that's hard to explain to newcomers is bad too. This is really difficult, and given the questions I've seen over the years here, I'm not even sure it's possible. Almost the entire set of poll questions is gray area. We can agree we don't want the ones that'll get 100 answers, and if there are really only going to be three, it's fine. But most are somewhere in the debate-prone middle ground.
I don't think there's a good solution to this currently, and even if we come up with something, we won't be able to handle the really popular things, like basics cookbooks. On top of that, some of the short-list poll questions can be written as good subjective questions anyway ("how can I learn vegetarian cooking with my kid?" could be good) so there's even less headroom.
Handling poll questions better
We can try figure out how to deal with questions with tons of answers, some of which came in an initial surge, and some of which are just one more getting tacked on three years later. This seems to be somewhere between difficult and impossible in the site's format. Our cookbook/resource type questions demonstrate this very well - duplicate answers, rambling answers, low-scoring excellent answers...
But if we had better tools, we could start to address this. Think of it as Community Wiki 2.0 - we still probably throw per-user rep out the window, grant people broader privileges, and think of it as building a single page. But we don't want a free-for-all. A couple main things come to mind:
a voting queue, to get enough votes on both new and old answers to put them in a reasonable order. Think Google Moderator - show people with voting privileges the answers that haven't reached a "consensus" yet. Additional upvotes on something already at the top don't mean much.
ability to merge answers. This is tougher than it sounds, probably involving something like the suggested edit queue, affecting revision history, and so on. (Duplicate answers are common, but partial dupes are even more common.)
The existing system is nothing like this, of course, and it's not something we can build ourselves - it's a task for the devs. But I think we've seen ample evidence over the years on SA, and before that on SO, that with the current structure, these questions really just don't work - we need something more.
Example questions
Your first example is a kind of hybrid of a decent on-topic subjective question asked in not quite the ideal way in the body (what approaches are good for getting started learning?) and an implied broad poll question in the title (what resources are good for beginners?). The subjective half is cool, but the poll question is pretty iffy. The broad spread of answers is a good demonstration that while we certainly do generate some helpful lists of things, it doesn't work well and we should be wary of allowing broad poll questions without changes to our system. (This makes it tough to make the right call on that specific question, because it has the good subjective stuff in it right alongside the uncurated mishmash.)
The second example is decent purely because it's a list question with a pretty limited potential set of answers. That's kind of a separate debate, and I think there are decent arguments on both sides. The big problem is that it's tough to draw a line between an okay restricted short list question and a huge open list question. It's unrealistic and often unfair to try to say something like "there will probably be five or fewer answers." I'd like to allow some questions like that (even within the current system), but I don't see a good way to define the scope, especially in an easy to understand way for newcomers.