Replication over slow connections

Hi

I was wondering if I could get some assistance regarding the setup of a
replication environment across a dedicated 64KB satellite link with an
average latency of about 800ms (occasional time outs).

To give you a bit of background, I have setup Transactional Replication with
one subscriber using the pull mechanism (with continuous synchronization).
The subscriber and publishing database servers are running SQL Server 2005. I
set the merge agent's default profile to "slow link" at the distributor.

I setup the snapshot agent to create a compressed snapshot to reduce the
size of the initial transfer. Once everything was setup I confirmed the
tables were all created at the subscriber once I started synchronization. I
left work and the next morning I expected to see the tables populated.
However they weren't.

I right clicked on the subscription and selected "View Synchronization
status" and found that the job had stopped with "Could not connect to
distributor". I thought that perhaps a few timeouts had caused this, so I
clicked "Start Job" at which point I got an error stating that the job had
already started. This did not make sense to me, if the job was running surely
it should be retrying? I checked the connection by pinging the remote server
and logging in via remote desktop. I left it for a few hours as I thought I
caught it at a point where the error had just occurred. Several hours later
the error was still there.

I took the following action, I stopped the subscription synchronization job
manually and then restarted it by right clicking "View synchronization
status" and then clicking "Start Job". It seems to be up and running again as
the status is still "Synchronization ins progress". I noted now that that
database is locked, I assume the data is now being transferred and bulk
imported? At the distributor I see log records on the "Distributor to
Subscriber History" screen through the replication monitor like the following:

Extracted file 'DBNAME_BV6e3b4288_4#24.bcp'
Extracting snapshot file 'DBNAME_BV6e3b4288_4#25.bcp' from cabinet file ...

Does this mean everything is in order? If it is, how do I prevent the issue
where I have to manually stop and start the synchronization job at the
subscriber?

Any help here would be appreciated.

Thanks
Rohland