warm standby server stops doing checkpoints after a while

Hello list,

I have a problem regarding running a warm standby server as described in
the postgresql 8.2 documentation.

I set up two servers. Both running PostgreSQL 8.2.3-1PGDG on Fedora Core
6 (x86_64). (Master driven by AMD Opteron / Slave running on Intel Xeon)

The master server copys its WAL-logs to the slave using rsync over SSH
which are then replayed by the slave using a little perl script which
does not really do much more then the pseudo code in the documentation does.
To this point everything works just fine.

To reduce space consumtion of the log archive my restore script checks
pg_controldata when it enters a wait cycle and deletes all log files
which are older than the last checkpoint segment. (Checkpointing happens
about every 5 to 10 minutes)

The problem is that the slave server stops checkpointing after some
hours of working (about 24 to 48 hours of conitued log replay).
After this point in time the slave does not do checkpoints anymore but
still replayes every log that it gets from the master. The master works
on as expected doing checkpoints every 5 to 10 minutes.

I raised loglevel to debug5 but there's nothing interresting in the
logs. Just log replay and the fact that there's absolutely no sign of
anything related to checkpoints.

Restart does not help. Only beginning with a new base backup brings the
slave to work again as expected.

Both servers are configured identically except archiving/recovery settings.

Does somebody has a clue what could cause this behavior of the slave server?
Could upgrading to 8.2.4 help? - I didn't find something related in the
release notes.
Could this be a bug?

Greetings,
Frank Wittig