Today I found out the rss feed that I wrote for the infobot to work out the current time was broken. Basically, the script goes off to timeanddate.com and looks at the tables on there to work out what time it is the world (since it's easier than working it out yourself - especially when countries keep updating their laws about what time zone they're going to be in and when they're going to daylight saving.) The script as read should have been able to parse the string, but we were getting all these "Tor"s in our output.
Of course, "Tor" is Swedish for "Thu".
Yep. The webserver was looking at the accepted languages and if there was no header it was sending back a sweedish version. So what I was seeing with my browser (that was sending those headers) wasn't the same as what LWP was getting. It never used to do this - to be honest, we English speakers of the world are too lazy, we just assume if we don't specify these kinds of things we're going to get English back.
An additon of:
# all those need to have the english headers now
# or time and date will send us the times in sweedish. bork bork bork
WORLD->header("Accept-Language" => "en");
US->header("Accept-Language" => "en");
EUROPE->header("Accept-Language" => "en");
and everything works again.