The ping= property does not have any relevance. If the ping fails (tested over HTTP where the webserver has been SIGSTOPed), the backend is not set in an error state. As a result Apache assumes that the bad backend is still healthy. Also I would expect the ping to test if the backend is healthy before sending the request to it. If the ping fails, the request must switch to another backend. But this doesn't happen (because the origin is not set to an error state).