What it did was accidentally ignore the fragment of the URL. The URL will still work, but the part after the # is a pointer for the web browser to tell it to go to a certain part of the page.

The # aren't pointers for the web browser here - that's what the bot assumed, though.

[ is the [ bracket (and ] is ]), which needs to be encoded for the URL to work with MediaWiki.

Another example.

&#91 is not valid URL encoding. Per the RFC the very first instance of the # is where the URL fragment begins. URL encoding is in the form of a %## where the numerals represent an ASCII code.

That might very well be true, however, we have lots of these non-standard-URLs and there are also websites using the # symbol for other purposes (Wikipedia, for example).

I believe that we need to find a way to deal with this problem, as it is clearly destroying links which are handled perfectly well by MediaWiki and standard browsers.

Fixed in beta10