(If anything Washington, which is facing roughly linear case growth without a shelter in place in effect, is arguably handling the best of anywhere in the US. But I recognize NYC being so dense is intrinsically in a bad spot in a pandemic)
I don't really know, but I'll guess. Maybe the estimate is done by taking the total for one day and subtracting out the total for the previous day? If the data were a bit messy and revised down, this might result in an occasional negative value.
Their testing rate is not even particularly high - Washington State has tested more per capita and has a drastically lower positive rate
https://covid-19.direct/state/NY
(If anything Washington, which is facing roughly linear case growth without a shelter in place in effect, is arguably handling the best of anywhere in the US. But I recognize NYC being so dense is intrinsically in a bad spot in a pandemic)