Could there be
a race condition
There could be race condition — maybe OS related. I was not able to reproduce the issue while logging and logging out rapidly and continually querying pg_stat_activity table (tested on FreeBSD). Most of the time I didn’t even see the process at all in pg_stat_activity and when I did it had the client_addr propulated.
Could it be that the process you are seeing is a parallel worker? That would not have client_addr populated as only the leader process has client_addr populated.
