"KG" == Kai Grossjohann writes:

 KG> Okay, this is the first time I've been bitten by MULE (at least I
 KG> presume that's where the error comes from).  I nnoo'd a new
 KG> backend, inheriting most of the stuff from nnmbox.  I created a
 KG> server for this backend.  I added a message with a Latin-1
 KG> character.  Now, when typing `g' in the group buffer, I get asked
 KG> (two times, I've got two groups): Revisit file literally?

 KG> When I say yes, I get that dreaded \201 thingy.

 KG> What do I do now?

unfortunately, this seems to be a hard-to-fix bug in Emacs/MULE. I
also "suffer" from this problem: from time to time, when i receive
mails to my mbox which contain some binary characters, this leads to
brokenness of Gnus' dealing with my mbox. Each time pressing `g' in a
group buffer starts to take a LOT of time (with bug mbox size, such as
10MB and more); note that if the mbox does not contain those `broken'
messages with some bytes which confuse MULE (or gnus?), pressing `g'
takes much less than a second (if there are no new messages). And
also, in this case article numbering in group buffer reported by gnus
becomes broken (commonly overestimated by 1 or 2). Lars some time ago
made the following patch for files.el which "hides" the Emacs'
questions "Revisit file literally?", but this patch ODES NOT SOLVE THE
REAL PROBLEM: the above-mentioned symptoms are still valid; and they
are TRIGGERED by the presence in the mbox of some `broken' messages (i
have some examples and could send them). So, i'd be grateful to you if
you could report this problem to RMS.