My impression is no one reads the tag wikis. And I am not sure references should be in the tag wikis anyway. Moving all our references into the tag wiki would make them explode due to the lack of pages. They'd more like dumps. IMO, tag wikis need an overhaul before we can consider moving more stuff into them. Also, we'd need to discuss whether frequently asked questions like in the reference question should go into a wiki at all, since technically they are questions with answers.
On Reference questions in general
Reference questions might not be a perfect fit for the Q&A format, but they serve a specific purpose. Compare the following:
1 - Reference - What does this symbol mean in PHP?
Solved a shortcoming in the SO search. Does not duplicate existing answers, but just links to them. IMO, the best format because it just organizes what is already available at SO. Of all the references, this one is the most successful (disclaimer: I created most of it, so I am biased).
2 - Reference - What does this error mean in PHP?
Solved the problem of questions asking for the same old errors people get in their code. There is a general answer for the most common error and links to existing q&a. Not optimal IMO, but for that particular purpose all right. It should be noted, that most of the questions this one closes are of the too localized "fix my bug plz" type where the OP is too lazy to search or understand the error message and just throws a wall of code at us.
3 - How do you parse and process HTML/XML in PHP?
A question hijacked and turned into a Canonical. This one had a couple of other questions merged into it, centralized most relevant answers into one page. While technically somewhat of a shopping question, this one was deemed necessary to combat the daily "how do I parse HTML with regex" and "how do I write a parser" questions. It's an entry point to the general topic.
4 - Reference: What is variable scope, which variables are accessible from where and what are "undefined variable" errors?
A self answered one. Unlike the others, this one is not a community effort. Personally, I don't like this one too much because it acknowledges that the question got asked previously, but then ignores any existing answers to it. Instead the OP provides one (thorough) answer himself, which leaves the impression that the OP wrote it mainly for his own good. I do not see why this required a separate question instead of hijacking an existing one as in Reference 3.
With the exception of the last one, all of these tried to organize existing content to fight common dupes. If these references help to cut down on these dupes, then they serve a purpose and - as long as we lack better tools to organize content - should be tolerated despite not being 100% real questions.
On the PDO Reference
I am not that familiar with the PDO tag, but I am sure there is a lot of repetition in that tag as well. However, IMO the reference suffers from a few problems:
1 - There is no links to existing answers
Basically the same problem as with Reference 4. If we already have existing Q&A on that, why not link to these? Chances are other people already provided good answers to these questions. A reference that doesn't take these into account, is taking away from previous people's effort. This is particularly bad when these are quality answers. IMO, the only reason to provide your own answers is when there is no good previous answers.
2 - The answers are inferior to a full page of Q&A
If you compare the answers with the amount of information you find on existing answers on the same questions, you'll notice that the existing Q&A contain much more insight into a problem. The reference does not look into edge cases or alternative approaches, effectively suggesting that the One True Way™ to approach the problem is in the given posts. This is not a problem in any of the other references linking to existing Q&A.
3 - There is only five answers
That makes the reference hardly worth keeping in my book. I mean, yes, the questions might get asked frequently, but the justification for creating a reference should be based on the amount of dupes it can help close. Given the problems outlined in 1 and 2, it would make much more sense to attempt to create Canonicals for these five questions like in Reference 3.
Tag wikis are insufficient to provide the same value as a reference due to the way they are implemented. References should be uncommon though and yield a justifiable benefit for not following our guidelines for acceptable questions. Large references should either overcome technical problems or provide a first line of defense against dupes in a particular narrow enough topic. When creating references, these should build on or include existing quality Q&A to make sure previous efforts don't get lost. Community created Canonicals are preferable to references since they do not have most of the problems references have and are generally much less controversial.
To prevent the repeated opening and closing of references and comment storms, I suggest anyone considering to open a reference in the future to ask on MSO first to get the approval (and help) of the community.