So Bryan and I were looking at the l10nupdate scripts, noticing that it's essentially a reproduction of what's in scap.
For a little work, it was possible to make l10nupdate quicker, but there's a lot of code in it that should be replaceable with a scap call.
When we started digging into this a bit further, we have come to a conclusion that any scap run after an l10nupdate is possibly wiping out any changes made by them. This is due to different cache directories being used to build l10n caches from... Causing untold bugs?