[Exim] Multiple delivery of messages

Has anyone seen this behaviour before? (exim version 2.02)

Someone sent mail to 1000 generated addresses, of which about 50% were
valid. The valid sddresses were successfully delivered:

2000-11-29 00:59:41 140qfA-00018P-00 => xxxxxxxx@??? D=cluster_redirect T=cluster_smtp H=imap5.herald.ox.ac.uk [163.1.0.213]

and the invalid ones failed:

2000-11-29 00:59:40 140qfA-00018P-00 ** yyyyyyyy@???: unknown local-part "yyyyyyyy" in domain "herald.ox.ac.uk"

All well so far. However, the delivery error message could not be
sent:

2000-11-29 01:00:15 140qfA-00018P-00 Process failed (1) when writing error message to zzzzzzzz@???

probably because:

2000-11-29 01:00:15 Error while handling error message: excessively long header line read

The original message remained in the spool. None of the recipients
were marked as delivered. The next time the queue-runner ran, the
message was delivered again, the delivery error could not be sent, and
the original message remained in the queue. This loop repeated until
our attention was drawn to the problem - by which time the unfortunate
recipients had received a dozen copies of the message.

Is this a known problem (exim version 2.02), and has behaviour changed
in the current version?

--
Ray Miller <ray.miller@???>
Unix Systems Programmer
Oxford University Computing Services