SQL 2005 Server (SP2) To SQL 2005 Express (SP2) Replication Looping/Excessive Repl Times

We have a large number of clients attempting to replicate two publications on 2005 Express databases (2 publications subscribed to the one subscriber database) with our 2005 Server (9.00.3042.00 SP2 Standard Edition) and experiencing two significant problems:

1) Users experience the following message:

The Merge Agent failed after detecting that retention-based metadata cleanup has deleted metadata at the Subscriber for changes not yet sent to the Publisher. You must reinitialize the subscription (without upload).

This problem should not apparently occur with SQL Server 2005 (or 2005 Express) instances with SP2 applied. All clients experiencing this problem have SP2 installed as does our Server and the retention period is 30 days. The subscribers have been replicating well under that.


2) Replications never succeed after appearing to replicate/loop around for hours

This issue is the most critical as we have clients who have been installed and re-installed with new instances of SQL Server 2005 Express, new empty databases (on subscriber before snapshot extraction), and using fresh snapshots (less than an few hours old) which cannot successfully replicate.

Interestingly there is at least 1 instance where several computers are subscribed and successfully replicating the same database as another where replication refuses to succeed.

To test we have taken a republished database from another 2005 Server which is working fine and restored it to the same server as the one holding the database with which we are experiencing problems and subscribed to it. This test worked fine and replication of both publications went through fast and repeatedly without showing any signs of problem.

This indicates that the problem is perhaps data related as it appears localised to that database.

Below are two screenshots which may assist.

Screenshot 1 Shows that on the server side the replication attempts look like they are succeeding despite the fact that the subscriber end does not indicate success. Also the history indicates the the subscription has spent all it's time initialising and not merging any changes.

Screenshot 2 Shows a rogue process which has appears on many of the problem child subscribers. It shows a process running with no end time even though the job indicates failure in the message and even though other replication attempts appear to have succeeded after it. This process stays in the history showing that it is running even when I can find no corresponding process for it.

Can anyone suggest a further course of action/further testing/further information required which may assist?

This is extremely urgent and any assistance would be greatly appreciated!

Thanks in advance!

Scott