I don't think the Sony Readers can handle plain raw HTML.
Why don't you make the whole thing into one massive ePub? I assume each wikipedia article is exactly 1 HTML file. Those are (possibly) fine as they are, though if the HTML is just scraped from wikipedia as-is, there might be a lot of markup that epub can't accept. Anyway, just treat each wikipedia article as a "chapter" when you create the table of contents. You can also make a multi-level TOC for easy browsing.
For example:
A
AA
AB
AC...
B
BA
BB...
Of course, even if all this is in one massive epub file, you can still search it.
Images can be included in the epub itself, shown in the actual article, just like on a real webpage.
Of course, each and every file and image will have to be listed in the manifest.
I would imagine it's a lot of work, especially depending on how the images are organized in the file. However, it might be easier if you write scripts to construct the OPF and NCX files, etc.
Just some ideas. =)