Bug summary

Plotting 2 trivial lines in a polar axes. Those lines are quarter-arcs with constant y values:

• line1 is only y=5
• line2 is only y=9
  The figure is plotted using a canvas.draw() call.

If line1 is deleted, set_ylim(8, 10) will make line2 disappear. In fact it can be found at y=13 for some reason.
Inspecting the object values (get_data) returns the correct values. It is as if the data was correct but the rendering off by just "4"

The issue is that set_ylim on a polar plot seems to apply an offset on the remaining lines after some workflow.

Code for reproduction

import numpy
import matplotlib.pyplot as plt


x = numpy.linspace(0, numpy.pi/2., num=90)
y1 = numpy.full(90, 5)
y2 = numpy.full(90, 9)

fig, ax = plt.subplots(subplot_kw={'projection': 'polar'})
ax.set_xlim(0, numpy.pi/2.)
ax.set_ylim(4, 10)
ax.plot(x, y1)
ax.plot(x, y2)
fig.canvas.draw()

ax.lines.pop(0) # If popping the 2nd line no issue encountered
if True:
    ax.set_ylim(8, 10) # This moves the 2nd line from y=9 to y=13

plt.show()

See simpler example at #24790 (comment)

Actual outcome

line2 rendering is done as if line2 had all its y values = 13

Expected outcome

line2 should be rendered as a if its values were y=9 (same as get_data return values)

Additional information

• Deleting the other line instead gives no issue. Only deleting a line that contains data inferior to every other lines triggers the issue
• Not calling canvas.draw before is not an option: My application is interactive the user is able to delete/add lines interactively
• The issue seems to come from set_ylim on a polar plot

Operating system

Debian 11.6

Matplotlib Version

3.3.4; 3.4.2; 3.5.3; 3.6.2

Matplotlib Backend

tkAgg; Qt5Agg

Python version

3.7.10; 3.9.2

Jupyter version

No response

Installation

Linux package manager