Please check and try out revision 7100. For example, with ipython -pylab:
x = np.arange(1000000, dtype=float) * 0.2
y = np.sin(x)
Then play around with panning and zooming.
To see what the behavior is like without the changes, just reverse the sign of x, since at present only monotonically increasing x is supported:
Notice that in the latter case, panning and zooming is jerky.