Firstly, I couldn't build the sample for some reason.
Secondly, we fixed the issue on our end, but there is still an issue with the library that might catch out other people. We figured out that something was going wrong with our calculation for the range of the DateTimeAxis and we were occasionally sending a range where the end time was before the start time - when this happened, the Axis clearly got stuck in some sort of loop and started eating memory. It really should bounds check the range, and error if given invalid values. We would have discovered our error much more quickly if it had.