The pgq queue is the backbone of a replication tool called Londiste. The idea of Londiste is to have a mechanism that is more simplistic and easier to use than, say, Slony. If you use Slony in a large installation, it is very easy for a problem on one side of the cluster to cause some issues at some other point. This was especially true many years ago when Slony was still fairly new.

The main advantage of Londiste over Slony is that in the case of Londiste replication, there will be one process per "route." So, if you replicate from A to B, this channel will be managed by one Londiste process. If you replicate from B to A or A to C, these will be separate processes, and they will be totally independent of each other. All channels from A to somewhere might share a queue on the consumer, but the transport processes themselves will not interact. There is some beauty in this approach because if one component fails, it is unlikely to cause additional problems....