What keeps me up at night? Nothing related to localization, although Client A's API reference often makes me wake up earlier than I'd like.

In five years we've come a long way in localizing it.

• Back when it was only a few hundred pages, Tech Pubs maintained it in FrameMaker and published it as a PDF. Localization was a headache because Acrobat played up with some of the characters in Chinese.
  
• It got too fat for that, so they dumped the PDF and went from FrameMaker to HTML to CHM using WebWorks. WebWorks was a localization nightmare, so I was glad when they outgrew that.
• Then they authored in RoboHelp and went straight into HTML/CHM. That was a little better, but Asian support in later versions of RoboHelp was elusive or costly or both.
• Now Tech Pubs doesn't do any writing in the API Ref. They maintain about a dozen Perl scripts for extracting HTML files (over 3000 now) from text-based header files, then compiling a CHM.

No matter how they cut it, it's a lot of work when it comes time to localize the puppy. I look for ways to streamline the process, but they all end up in the tail wagging the dog, because L10n is so far downstream and at the sorry end of the food chain. Besides, most any improvement would require endless, largely fruitless nagging of a platoon of engineers, all to save a few thousand dollars a year.

There's ROI to localization, but not necessarily to every facet of localization.