The attached patch appears to dtrt in my tests; I've not applied it
yet as I'd appreciate another set of eyes to check if there's
anything
obviously broken. I've also attached the difference in the
excuses.html generated by this morning's britney run and a test run
with the patch applied and the same input data - the output is now
more noisy in the case of source to binNMU dependencies (e.g. the
haskell-wai-eventsource entry at the end of the excuses list) but I'm
not entirely sure that's a bad thing.