Discovery: what saturation metrics for Postgres are we not tracking as part of the saturation monitoring framework and Tamland?

As part of &474 (closed), @brentnewton has requested that we, as a department, improve our understanding of the key resource saturation bottlenecks in our Postgres infrastructure.

As a first step, as a way of starting this process, it might be worthwhile to explicitly state all the different ways that we think Postgres can fail.