How to make an Axes transparent to events? - python-3.x

Question

I am trying to find a way to make an Axes object passthrough for events.
For context, I have a figure with 6 small subplots. Each of them responds to mouse motion events by displaying a cursor dot and text info where the user aims. I also made it so that clicking a subplot will make it as large as the figure for better visibility. When moving the mouse over invisible axes, event.inaxes will still point to that ax despite being set to invisible and that is what I would like to avoid.
Below is the MRE:
import matplotlib.pyplot as plt
fig, (ax1, ax2) = plt.subplots(1, 2)
ax1.set_gid("ax1")
ax2.set_gid("ax2")
fig.show()
def on_movement(event):
"""Write on the figure in which `Axes`object the event happened."""
width, height = fig.bbox.bounds[2:]
x = event.x/width
y = event.y/height
text = ax.get_gid() if (ax := event.inaxes) is not None else "None"
fig.texts[:] = []
fig.text(x, y, s=text, transform=fig.transFigure, c="white", bbox=dict(fc="#0055AA", ec="black"))
fig.canvas.draw()
fig.canvas.mpl_connect("motion_notify_event", on_movement)
As expected, as you hover the mouse over ax1, the empty gap and ax2, you will see one of those three texts appear:
ax1.set_position((1/3, 1/3, 2/3, 2/3))
Same thing as I arbitrarily resize and move ax1 so that it is partly overlaps with ax2.
ax2.set_visible(False)
Now this is my problem. Invisible axes still trigger events. Is there a way to make some axes "transparent" to events? Obviously the usual technique of sorting all the cases in the callback does not work here.
Currently envisaged solutions:
ideally, finding a setting akin to zorder so that the "highest" axes gets the event.
ugly workaround: set the position of the invisible axes to ((0, 0, 1e-10, 1e-10)).
less ugly: working with figure coordinates to convert event.x, event.y into event.xdata, event.ydata for the only ax that I know is visible. Basically xdata1, ydata1 = ax1.transAxes.inverted().transform((event.x, event.y)) if event.inaxes is not None + see if there are edge cases.
The latter is already implemented and does work, so save your time if you want to write a reply using that approach. I'm mostly interested in an amazing one-liner that I would have missed, something like ax2.set_silenced(True).
Python 3.8.5
Matplotlib 3.1.3

Answer 1

Well, setting the appropriate zorder does work actually.
ax1.set_zorder(2)
ax2.set_zorder(1)
...
def on_movement(event):
...
fig.text(x, y, ..., zorder=1000)
...