I was torn whether to post this to a SQL CE or the SQL Server Replication forums, but all my attempts to isolate this failure mode in test cases using pure ADO.NET against SQL CE have been unnsuccessful, so I'm leaning toward this somehow being related to
an interplay with merge replication.
Some of this background may not relate to my specific failure, but I'll err on being thorough: My topology is SQL Server 2008 R2 running Merge Replication with Web Synchronization using the SQL Server Compact Edition 3.5 SP2 Server Tools, and SQL Server
Compact Edition 3.5 SP2 for the client DBs on full Windows XP or later OS’s. We have 150+ articles, each of which range from 10 published columns to close to the 255 limit. Regarding filtering, we have a root article filtered on the hostname, which
then has 2 levels of join filtering cascading off that. We are using column level change tracking and with use_partition_groups=false (due to a major performance hit when doing DML updates on the server when this was "true") and we are not pregenerating
snapshots for specific subscribers so that occurs on the fly for new subscribers. Currently we have subscribers counted in the hundreds, but the end target will be thousands of subscribers across the US synchronizing over the public internet (including
from air cards).
Now for my sp
View Complete Post