Jeff,

The error codes in your message are pointing to a bug in replication to SQLCE.  There is a workaround, I found it in Knowledge Base article 300597. Try this workaround...

15 GB is not that big for merge replication, we are replicating tables in merge replication much larger and they run fine.

To eliminate the query timeout errors, set the -QueryTimeout...

Jeff,

You should not be getting this behavior from SQL Server merge replication.  I have a sneaking feeling that the snapshot on your merge publication is set to delete all data...