It appears that there has been some Microsoft auto-update that has occurred within this month (November 2007) that has caused a .NET 2.0 application that I wrote (using C#) to convert A0 to EF BF BD when running the following code: Previous to Nov 3, 2007 this application did not convert A0 and all was well. A0 happens to be a   in ISO 8859-1.
There's always a "g", "s", and "z" parameter, and sometimes a "o" parameter that has values of 2 or 3. We've noticed that with our pages, a certain version of this script is correlated with seeing corrupted characters in the DOM: if "o" is omitted or set to 2, we'll see a Unicode FFFD injected near the end of the page or sometimes a Ux000E ...
Wenn nun bei dieser Kodierung ein Zeichen gefunden wird, welches nicht im Unicode abgebildet werden kann, so erhält dieses Zeichen den Wert EF BF BD – besagter Replacement Character. Damit ist dann auch erklärt warum die binären Dateien hauptsächlich nur noch aus diesen Zeichen bestanden. Die echten Daten wurden beim Kopiervorgang ...