Obstacles remain in costmap despite observation_persistence being set
My question is mostly a duplicate of this one but since it was closed without answer long ago, I open a new one.
I expected that when observations are removed after the given time, the corresponding obstacle should disappear from the costmap. However, this does not happen, the obstacles stay there forever.
Is this a bug and the observation_persistence parameter is not working as intended? Or is it expected behaviour? In this case, what is the purpose of this parameter?
In any case: Is there a way how I can achieve my desired behaviour (i.e. obstacles disappear after some time from the costmap when they are no longer supported by current observations)?