Since upgrading from PHP 5.2.0 to PHP 5.2.2, I have the very same problem. I'm using the latest versions of WordPress and WPMU and the problem occurs on both WP and WPMU.
The characters don't show up visibly in all browsers, but they are there on every page on every WordPress blog on my server in the raw HTTP output.
See, for instance, the '1f48' right before the page output and the '0' at the end of the page in the example below:
GET /weblog/feed/ HTTP/1.1
Host: zoetermeer.sp.nl
HTTP/1.1 200 OK
Date: Mon, 07 May 2007 20:33:12 GMT
Server: Apache/2.2.4 (Unix)
X-Powered-By: PHP/5.2.2
Last-Modified: Mon, 07 May 2007 11:22:54 GMT
ETag: "ca3335e1a2642f1c67458017a4655e53"
Status: 200 OK
Transfer-Encoding: chunked
Content-Type: text/xml; charset=iso-8859-1
1f48
<?xml version="1.0" encoding="iso-8859-1"?>
<!-- generator="wordpress/wordpress-mu-1.0" -->
<rss version="2.0"
xmlns:content="http://purl.org/rss/1.0/modules/content/"
xmlns:wfw="http://wellformedweb.org/CommentAPI/"
xmlns:dc="http://purl.org/dc/elements/1.1/"
>
(...)
</item>
</channel>
</rss>
0
This breaks RSS parsing in some clients, and in some browsers the characters show up (strangely enough, they don't in most - I'm not seeing anything in Firefox 2 or IE 7)
The character (hex?) strings change when the page content changes (for instance, when I change the encoding from iso-8859-1 to utf-8). This leads me to assume that it has something do with character encoding, but I really can't find it out where the problem lies...
-Kevin