I can't explain it, two identical sites, as far as I know, and one of them constantly makes erroneous markup. A non-printing character, which others have identified as 0x04, <EOT>, gets inserted as the last, or sometimes first char of a file. Either place, or ones shuffled to the middle by edits, cause problems, with umbraco. Most applications can handle the characters, so just loading the files means you have removed them from the displayed text. The only way I've come across to confirm their existence is to load the file in an editor such as umbraco GUI or PLESK text file editor also, Ctrl+A copy all, and paste into Notepad++ which will render each NPC as a ?.

None of that matters, much, as I can only replicate on my host on my domain, but the end result is an EOT at the very beginning of umbraco.config, before the declaration. When Lucene or some other component sees this, it causes a crash, possible unpublishing of the site, and definite restart of the application as often as twice a minute on an entirely idle site.

Anyway, what I need is a brute force way to eliminate these characters, so I can know or not if they are cause of my problem. I have edited closely every text file, from xslt, to master, to css, etc. to content. The EOTs are not coming from me.

If I were to edit the source which I am unfamiliar with and can't even get to build usually, but I am motivated, where in the source does it write umbraco.config? I want to Trim the output. Please help.