The bug was probably actually yesterday, but in any case, they now have 8/23/2016 misdated to 8/24/2016, which is resulting in two rows for 8/24/2016. I've emailed their front-line support email about the problem, no reply yet.
I am not sure how this affects downstream Pipeline/Data sets, but it affected me this morning via fetch_csv
failing with a reindex to duplicate index bug (perhaps in fetch itself, perhaps during my pre-processing).
FYI.