The following bug has been logged online:

Bug reference: 1425
Logged by: Jonathan Pool
Email address: pool (AT) utilika (DOT) org
PostgreSQL version: 8.0.0
Operating system: Red Hat Enterprise Linux 3
Description: False failure message on boot start
Details:

When /etc/init.d/postgresql (as included in the RPM version of 8.0.0) starts
PostgreSQL as part of the host's boot sequence, it performs the -f test on
the PID file $PGDATA/postmaster.pid and tests whether the process has a PID.
The PID test passes, but the PID file test fails, despite the fact that the
PID file actually exists. So, the script reports a failure, showing up as
"FAILURE" in red instead of "[ OK ]" in green on the console. If this script
is run manually with a "restart" argument after the boot, the PID file test
passes.

I expected the PID file test to pass during the boot sequence, just as it
does during manual execution of the script, and thus to get an "[ OK ]"
report.

---------------------------(end of broadcast)---------------------------
TIP 8: explain analyze is your friend