RPA.Assistant dialogs causing 100% CPU after close

[kgraefe]

After closing a dialog Robot continues as usual but in the background one additional thread is spinning at 100% CPU. Opening and closing more dialogs creates more spinning threads (one per dialog).

To reproduce, run the following minimal example, close the dialog and watch htop or similiar during the Sleep:

*** Settings ***
Library  RPA.Assistant

*** Test Cases ***
Hello World Test
  Add submit buttons  buttons=Confirm,Failed  default=Confirm
  Run dialog    title=Success
  Sleep  300

I first encountered the issue on a Debian Bookworm container running on my Fedora 42 box with forwarded XWayland. In this container we run robotframework==5.0.1, rpaframework==27.7.0 and rpaframework-assistant==3.0.6 (because for now we need the option to fall back to RPA.Dialogs). In that scenario I tracked down the spinning thread to spin at https://github.com/flet-dev/flet/blob/v0.4.2/sdk/python/packages/flet/src/flet/sync_local_socket_connection.py#L146 with py-bt in gdb showing Waiting for the GIL.

However, I can reliably reproduce it on my box (without container) using the latest versions (robotframework==7.3.2, rpaframework==31.0.0, rpaframework-assistant==5.0.0) with both, Wayland and XWayland. Only I could not get py-bt to run in gdb, but I guess it probably is the same issue...

[kgraefe]

I can reproduce on my Fedora 43 box as well (I needed to downgrade Python, though, as it ships with Python 3.14 which is too recent for RPA):

❯ sudo dnf install python3.13 python3.13-devel
❯ virtualenv --python=$(which python3.13) venv
❯ . ./venv/bin/activate
❯ pip install rpaframework==31.0.0 rpaframework-assistant
❯  # copy minimal example from above
❯ robot test.robot
# confirm dialog, observe htop in a different shell

Edit: same on Fedora 43 with Python 3.9.25

[kgraefe]

Same on a fresh install of Debian 13 Trixie. :-\ Is there sth. wrong with my test case? Do I need to shut down the dialog or sth.?