Comment 6 for bug 2063353

This is a difficult one to reproduce consistently, but it is most certainly an issue.

I had a relatively fresh installation and went through the following to get it to happen (meaning it worked every time until the last line):

plasma-discover --backends packagekit-backend
plasma-discover --backends snap-backend
plasma-discover --backends snap-backend,packagekit-backend
plasma-discover --backends packagekit-backend,snap-backend
plasma-discover
plasma-discover
plasma-discover --backends packagekit-backend,snap-backend

Using a live ISO, I made a little script to continuously run through that pattern and was actually able to get it to fail eventually. I found that the failures only occurred with the packagekit-backend,snap-backend call.

I then modified the script to only do this one call and keep track of the total number of tests and total number of failures (as tested by checking to see if the window was open by grepping for Discover in `wmctrl -lp` after a sensible wait time) and was able to reproduce with some degree of regularity.

In particular, after 300 tests, I had amassed 113 failures, amounting to a 37.7% failure rate.

Definitely a race condition seemingly related to threading if the upstream bug is to be believed.

This is with Lubuntu 24.04 in an X session and a generic kernel, I might add.