I saw I only had Native Client 2008 installed on the Server - installed 64bit Native Client 2012 from (https://www.microsoft.com/en-us/download/details.aspx?id=29065)

NB: Download link is found under INSTRUCTIONS NOT when clicking download.

Will reboot and see if there are any changes.

No luck :(

However now when some workspace fails, it is in 90% of the cases 2 of the particular workspaces. I suspect that there are just too many joiners / sqlexecutors / readers either in one workspace or that the connection after them does not get cleaned up properly after each run. Unfortunately it is difficult for me to get access to the admin of the database to see the connections. I will try that today and keep this thread update for anyone else with similar issue.

It seems the issue is due to the process taking to long to process - expiring the timeout for keeping the database connection. even though it was set on 10 minutes on all joiner, reader, sqlexecutor. The trick was to optimize the workspace and replace sqlexecutors with joiners (with Prefetch-queries). We only worked with 8000 rows in this case, however it really slowed down due to very non-optimal workspace. For instance sending 8000 sqls one and one to a server, and doing this multiple times resulted in runtime of more than 10 minutes. With optimizations it seems to work great and spend less than 1minute in the same data.

Perhaps write an error in the log when "Database connection from Joiner_4 / FeatureReader_2 / SQLExecutor_2 have timed out (timeout set to 600) - optimize query or set a larger timeout) - so you can easily see in the logfile - and find the correct Transformer among many

In the fmeServerConfig.txt there is a part:

#--------------------------------------
# Limits
#--------------------------------------
MAX_TRANSACTION_RESULT_SUCCESSES=100
MAX_TRANSACTION_RESULT_FAILURES=10
MAX_FAILED_TRANSACTION_REQUEST_RETRIES=3
MAX_REGISTRATION_READ_ATTEMPTS=6000
REGISTRATION_READ_RETRY_WAIT=10
MAINTENANCE_IDLE_WAIT=90000
TRANSACTION_DB_RETRY_WAIT=20

1. It this only related to communication between FME Engine and Core - or does it also apply to any other databaseconnections?

2. Is there a parameter that defines the maximum number of databaseconnections from an FME Engine? Some simple math if a assume one FME Workspace that contains 10 reader/joiner/sqlexecutor with the same database connection uses 10 database-connections (1 connection per transformer) and they run 10 workspaces on 10 minutes. Then the number of database-connections to the same database by FME is 10 connections/workspace * 10 workspaces run = 100 connections per 10 minutes. 

3. Does one FME Workspace handle 1 Database Connection as a single connection, or multiple connections when used in several SQLExecutors,joiners,readers,writers? If FME handles one Database Connection as 1 connection through the whole run of the workspace (regardles if there is a Joiner, SQLExecutor, Reader, Writer) that uses the same connection - then the needed connections is only 10. 

Is there any chance the link between FME Server and the Database system drops at all? Even for a split second? I'm really curious to learn more about this. You have experienced several variations of the SQL Server provider error.

We are fairly sure now that this was a known issue with the Joiner which was reported and fixed in FME 2016.1.2 Build 16670 (PR#71929). The issue is known to have caused workspaces to fail intermittently on FME Server when the Joiner was connecting to Microsoft SQL Server or Oracle and possibly other databases. We will create a Knowledge Center article shortly and include the link here. Upgrading FME Desktop and FME Server to 2016.1.2 or later and upgrading any Joiner transformers within the workspace should hopefully resolve this issue.

I tried to replace all Joiners, and I also forced these group of workspace to use only 1 specific engine. I still have workspaces using both joiner, reader, sqlexecutors - however now they do not seem to fail. I suspect there is an issue when 2 different Engines runs simultaneously connecting to the same MS SQL-database.

Well, case solved for now, and I`ll repost in this thread if the issue starts again. Thanks for all the reply - and the reason I put it here on the forum is for people to easily try the different solutions here if they experience issues with MS SQL on FME Server.