This should now be fixed (http://mail.python.org/pipermail/python-dev/2013-January/123680.html).

> This seems to happen whenever the revision description begins with
> text that results in a link to something else (e.g. an issue number).

To elaborate a bit on what I wrote in the email, I'll provide a few examples.  Here <a rev></a> is a link to the revision page, <a bpo></a> to the bug tracker, and the 4 steps shows the original message, the resulting link in the "shortlog" page, the link added by interhg and how the browser interprets the invalid HTML:
1) issue at the beginning:
#12345: fix foo  <- commit message
<a rev>#12345: fix foo</a>  <- normal link
<a rev><a bpo>#12345</a> fix foo</a>  <- interhg (invalid html)
<a rev></a><a bpo>#12345</a> fix foo <- resulting html

2) issue in the middle:
Issue #12345: fix foo  <- commit message
<a rev>Issue #12345: fix foo</a>  <- normal link
<a rev>Issue <a bpo>#12345</a> fix foo</a>  <- interhg (invalid html)
<a rev>Issue </a><a bpo>#12345</a> fix foo  <- resulting html

In case 1) the revision link ends up being empty, and there's no way to get to the revision page.  In case 2) everything before the issue number links to the revision page, the #12345 links to the bug tracker, and everything after that is not a link.

> In contrast, the "graph" page seems always to link to the revision

This seems to be because that page is not affected by the interhg extension.

> the UI doesn't make a distinction between the portion of the 
> description that links to the revision and the portion that links
> to the something else.

That's because the link to the revision page ends right where the one for the issue begins.

> For example, with description text "Fix for fcc629208842",

FWIW all that I said about issues is true for changeset ids.  However the fix I added for issue has not been applied for csids (they are not really common in commit messages, and usually they are not at the beginning).

> It might be better if the revision link was separate from the
> description text.

I did the opposite -- I left the revision link there and created a separated link to the issue at the end of the description. 

I think this issue can be closed now, even if there still some room for improvement.