Understanding Postgres Memory Usage

I have an application that uses Postgres 9.3 as the primary datastore.  Like any real-life application, it's not all roses—There are many ugly, convoluted, and inefficient queries.

Some of these queries use quite a bit of memory.  I've observed a "high-water mark" behavior in memory usage:  running a query increases the worker memory by many MBs (beyond shared buffers), but the memory is not released until the connection is closed.  For example, here's the memory usage on my test server when running a query once and leaving the connection open.

$ free -h  # Before the query

             total       used       free     shared    buffers     cached

Mem:          7.8G       5.2G       2.6G       212M        90M       4.9G

-/+ buffers/cache:       248M       7.6G

Swap:           0B         0B         0B

$ free -h  # After the query

             total       used       free     shared    buffers     cached

Mem:          7.8G       5.3G       2.5G       212M        90M       4.9G

-/+ buffers/cache:       312M       7.5G

Swap:           0B         0B         0B

$ sudo -u postgres smem -trs uss

  PID User     Command                         Swap      USS      PSS      RSS

 8263 postgres postgres: postgres souschef        0   200204   203977   209540

 8133 postgres /usr/lib/postgresql/9.3/bin        0    50456    61090    74596

 8266 postgres /usr/bin/python /usr/bin/sm        0     5840     6261     7460

 8138 postgres postgres: autovacuum launch        0      776     1146     2968

 8139 postgres postgres: stats collector p        0      300      470     1872

 8135 postgres postgres: checkpointer proc        0      148      342     1880

 8137 postgres postgres: wal writer proces        0      140      322     1812

 8136 postgres postgres: writer process           0      132     6814    15140

-------------------------------------------------------------------------------

    8 1                                           0   257996   280422   315268

This is proving to be very troublesome on my production server because I use connection pooling (so connections remain open indefinitely) and the connection memory seems to rise without end, to the point where 25 open connections OOM'd a 4GB server.

So I have a couple questions:  Is this high-water mark memory behavior expected?  If so, how can I identify the queries that are using lots of memory and driving the high-water mark upwards?

I understand that this post is rather vague, I didn't want to talk your ear off with specifics in case this was pretty basic, well-understood behavior.  If necessary, I can follow up with an email diving into the specifics of what I'm observing.