Well, this would not solve the issue described in this thread, however "in the long run" it would cause much pains known as agent management hell... "install it and be done with it" is nothing but wishful thinking
Our current approach is much more reliable and usable, trust me on that one (in my previous software company, it was all about persistent agents - so I've had a chance to see how this is working for tens of thousands of customers).
Now, if there is something special about certain environments that make our in-guest process "hang" and not exit properly under certain circumstances, we would definitely like to investigate and address this (logs from guest VM will be requiring). Making it a persistent agent is not a solution, because persistent agents "hang" just as well as non-persistent ones