The calculations used in SubdividedQuantityRange and the transformations it uses breaks down due to lack of precision when zooming in deeply, far away from the origo. This typically happens with timestamps (using a 1970 reference), when pixels are around nanosecond size.

A consequence of these break downs are various glitches. For example thrown IndexOutOfBoundsException and similar, often with negative indices. Missing tick labels, grid lines, highlighting, and similar have also been observed in the past.

Sometimes glitches appear at a certain zoom level but disappears when zooming in slightly further. They reappear when zooming in even further.

While some of these calculations could possibly be improved to eliminate the temporary glitches, there's a fundamental limit that we would hit shortly thereafter anyway.

Thus, and since there's currently not that detailed data, we should simply limit the zooming depth in some suitable way. (In the future, JMC-3571 might make this problem easier to work around by using a different reference point.)