Which curve fit is best (if any)?

It's remarkable how once a big team starts working on a big hotlist of bugs (July '17 in this case), the progress rate is pretty consistent, even though the variance of time it takes to finish any one bug is surprisingly high. That's why in aggregate, it makes sense to try to estimate release milestones even though it's almost totally pointless to try to estimate the completion time of one or two bugs.

The red line assumes a linear rate of progress. The blue line is a much closer fit to the existing data, but assumes a quadratically increasing rate of progress.