plperl doesn't release memory

I have been experiencing an issue with plperl and PostgreSQL 8.0.1 in
that after calling a plperl function memory does not get released. Two
different systems and each show different symptoms:

1) system: FreeBSD 5.3-Stable i386 with 1 GB RAM, dual Xeon P4
processors.
   script: plperl issues an SPI_EXEC_QUERY('select rows from table where
condition');
this loads the result set into memory (to the tune of some 600MB based
on top output). The function iterates through each row to grab some
totals information and spits back a number.
On the 2nd iteration of this function the connection is lost :

Out of memory during request for 1012 bytes, total sbrk() is 291207168
bytes!
Callback called exit.
LOG: server process (PID 12672) exited with exit code 12
LOG: terminating any other active server processes
LOG: received immediate shutdown request
LOG: all server processes terminated; reinitializing
LOG: database system was interrupted at 2005-03-23 17:17:23 EST
LOG: checkpoint record is at 2/4D7F206C
LOG: redo record is at 2/4D7F206C; undo record is at 0/0; shutdown TRUE
LOG: next transaction ID: 4913594; next OID: 60798748
LOG: database system was not properly shut down; automatic recovery in
progress

2) system: FreeBSD 5.3-Stable amd64 with 8GB RAM, dual opteron
processors
   script: plperl issues an SPI_EXEC_QUERY('select rows from table where
condition') which fetches roughly 18k tuples of rather small size. Each
row is acted up and if criteria are met, a reference to the row is
pushed onto an array (reference).
   after several iterations of this script (a dozen or so), a file is
COPYed into the database consisting of some 38k rows and each row is
acted upon by a trigger (plpgsql) -- this process normally takes just
under a minute, but after running the plperl function a dozen or so
times, the run time for the COPY exceeds 3 minutes. Restarting the
PostgreSQL backend (restart) brings the COPY time back down to
sub-minute range.

Is it normal for plperl to *not* release any memory? Or perhaps plperl
is not pfreeing or SPI_FINISHing cleanly?

Sven Willenberger