I have a morbid fascination with the flaws and shortcomings in Wikipedia process that lead to the website being held unreliable—even by itself.
Bystander effect is unchecked edit
The bystander effect is a phenomenon discussed in social psychology in which individuals faced with something wrong opt to do nothing rather than something. While Wikipedia's meticulous revision histories make it easy to analyze the instances in which editors opt to do something on Wikipedia, it is nearly impossible for the editor community itself to analyze the instances in which editors opt to do nothing. This is partly due to the site's privacy policy, which does not allow it to divulge data about what pages each user merely visit without subsequently editing them, but also more generally because we simply can't know what each user is thinking when they visit; whether they detect a problem or not, and what actions they take as a consequence, if any.
This in spite of the fact that the site is crucially dependent on a critical mass of editors doing something, and as such upon understanding when and why they instead choose to do nothing.
The vocal minority decides edit
Disputes over content in Wikipedia are decided through a number of different modes and channels, such bold-revert cycles, edit warring, talk pages, project pages etc. But of the millions of people who come to Wikipedia to read its content every day, only a small fraction participate in editing it through those channels. This is particularly the case in the most obscure topic areas, such as "villages in Uzbekistan".
As a corollary to the section above, we do not know specifically who opts not to participate through these channels, or why, or when. As such, it is difficult to determine what impact it has—not only overall, but also in individual disputes on specific pages. It is a "known unknown", and leads to an unyielding ambiguity in the outcome of disputes, which can be almost worse than definitely wrong information, which can nearly always be easily removed.