PITR recover ends unexpected

Hi,
since some month we use PITR and it works fine. We make a base backup
once a week. For testing if everything will work when needed we replay
the wal files every day to the base backup on a second machine. It works
fine except for two time when we saw a strange problem. The recovery
process stated a zero length record in the wal file and stop 'as
expected' as it told us. The recovered database was useable but we had a
gap in the recovered data from the time it stops replaying wal files up
to the next base backup. We didn't see any hardware problems or problems
while copying the wal files to the second machine. So our question is if
anybody out there saw a similar problem.

Thanks in advance for your replies

Bernd Rieke