When a QQuickWindow is hidden, we will end up calling invalidateGraphics() twice on the scenegraph.

The second time this happens, we will call makeCurrent() on the context which may at this point be for a non-existent surface. Typically, it will just leads to the makeCurrent() returning false, but as seen on Wayland (see QTBUG-88277) it is possible that the makeCurrent() succeeds if the driver supports the surfaceless context extension.

The issue on Wayland has been successfully worked around by detecting the case, but the core issue is that we are doing a redundant invalidateGraphics() on an already-hidden window.