In the ongoing effort to reduce the memory footprint of system level tools (CRI-O/conmon), there is the desire to change the architecture of how conmon is run. currently, there is a conmon process per container, as well as per container exec process. While this is simple, it has needless overhead (especially because of how go processes hold onto memory that they use).

As a openshift user, i'd like the infrastructure keeping containers running to have as small a footprint as possible. Moving to one conmon per pod will help with this