A better stop

Greetings

I was/am having the intermittent starting of postgresql problem. The error
messages recorded in the logs seem to imply some sort of shared memory
problem. The system is as stripped as possible, (only essentials and
postgresql) giving me the impression that a lack of buffers etc unlightly.
So I simplified my tests and came up with the following:

Postgresql will always start if I do not query the DB (using ODBC on an NT
server). If I shut down the server while it is in the middle of a large
query then it will not start next time round. The stopping of postgresql
reports failed. The shutdown command used 'shutdown -r now'. When shutting
down the '/etc/rc.d/init.d/postgresql stop' is run. There the killproc in
this script.
I am using the script out the box so to speak, the only alteration is that
of the logfile. It has all the cleanups that the mail list archives mention.

I find it difficult to believe I am the only one experiencing this. It seems
to me, and I am far from home ground here, that when shutting down the first
step should be to stop any incoming queries and complete any existing work
and then kill the process (perhaps this is what kill proc does).

Perhaps the shutdown script should change so the if the killproc fails it
wait a few seconds and retries. Again if it fails it should waits again and
again until it does stop successfully or a timeout is reached.

What are your thoughts on this? Am I going completely in the wrong
direction? Would changing the script be a good idea?

Regards
Warren

Config
Redhat 6.2 with PostgreSQL 7