I dug into this further, and believe I know what's causing this: an HTTP client is submitting a URL to the server which is not properly URL-encoded. Specifically, there is an invalid character in the URL.
To repro, paste the following at the end of your URL into IE8: default.aspx?p=´
If you examine the bytes going over the wire (e.g. using Fiddler), you'll see an actual Hex B4 character is being sent from client to server in the URL. This is an illegal character in a URL, since URLs are limited to char codes under 0x80 (any larger-than-0x80 char codes must be percent-escaped).
So your client is passing in an invalid character, and your server is (correctly) replacing the bogus character with %EF%BF%BD which is the UTF-8 encoding for the Unicode Replacement Character (U+0FFD), which is what happens when a character is encountered which has no equivalent in the local encoding.
AFAIK, this is a bug in IE. If you type the same URL into Firefox, Firefox will encode the URL properly (as %b4 instead of ´). Note that, also AFAIK, the problem only happens when manually pasting invalid characters into IE's address bar-- if the same character is present in a link, IE seems to encode the URL properly (at least in the cases I tested).
So you should figure out who is sending this bogus URL to you, and tell them to start encoding their URLs properly!