When an appender uses the "utf-8" encoding, output logs will append a BOM to the start of a file. This is problematic when the log file is only supposed to output errors. Explorer reports the file size as non-0, so the only way to know if an error actually occurred is to open the log every time and check. This is not desirable.
Further, from the Unicode FAQ (http://unicode.org/faq/utf_bom.html), a BOM is unnecessary for UTF-8 since it is already a sequence of bytes, so there are no endian problems to solve in the first place.
Removal of the BOM is not necessary, but should also provide a means to use UTF-8 without a BOM. That could be a new encoding name such as "utf-8_nobom" or as a separate attribute such as "bom=false" or something else.