I fully agree that we have a problem here, but don't think we're that far off a workable solution: technically, wikis can do the filtering already today — it is just that culturally (being so accustomed to the journal article as the container of new research), we are not ready for this yet.
Example: If someone were to have investigated whether castration would induce behavioral changes in guinea pigs, they could simply note this on the talk pages of the articles Guinea pig, Castration, Behavior, Hormone regulation or related articles, and include into their post a link to the relevant entries in their notebooks.
Of course, we could think of them putting the note simply on their blog, tagging it, and the community of those who watch the wiki articles linked to those tags (or categories) being notified automatically (e.g. via those talk pages, or mailing lists or Friendfeed groups).
In either case, the post-publication peer review would be about whether and how any of these wiki articles should be updated in light of the new evidence.
The idea of posting the news on talk pages serves the purpose of limiting self-promotion within the framework of current wikis, but if the wiki had a karma system (ideally at both article and systemic and possibly intermediate levels), researchers could update the relevant articles themselves, giving them the chance to increase their karma at the risk of karma losses if either the content or its presentation are judged non-appropriate by the relevant articles' communities.
Note added after posting of the comment: An example where an approach very close to the one described above can already be observed in practice is provided by the Polymath project: The edit histories of its Problems 4 and 5 reflect how knowledge on the topic has advanced in the course of the investigation.