+------------------------------+
| NodeA                        |
|  ??/15 x S1 -- 3011  <----------> 
|  ??/15 x S2 -- 3012  <---------->
|          S3 -- 3013 disabled |...
+------------------------------+  .
                                  .
+------------------------------+  .
| NodeB                        |  .
|  12/15 x S1 -- 3011  <----------.-->
|  15/15 x S2 -- 3012  <----------.-->
|   4/15 x S3 -- 3013  <----------+
+------------------------------+  |
                                  |
                                  |
                    +-------------|------------+
                    | NodeC       |            |
                    |             V            |  
                    |         Application      |              
                    +--------------------------+

External nodes can connect to NodeA and NodeB.
On NodeA and NodeB:

Ports 3011, 3012 and 3013 are equal but the processname:

SET SERVICE/PORT=<PORTNR -
	/file=<service.com> -
	/user=<service_user> -
	/PROCESS=<service_nr>
	/LIMIT=15

Any process starting on that service will first rename itself to hold
the BG-device ("S1" will become "S1_BGxxxx") to prevent name clashes.
This is part of <service.com>
The program that is run afterwards, will read the IP-socket, process the
message and return the outcome. After that, the service remains active
and will wait for the next request. So it is possible to handle up to 15
requests simultaniously (as far as the data-processing allows)

If a request is received on a port, and no service is started
yet, it will be created. It will read the message, process it, return 
the information (over the same port) and listen for the next request.
If a request is received on a port, and a service-process is listening,
this process will handle it: Read the message, process it, return the
information and listen for the next request.
If a request is received on a port, and no service-process is free for 
handling, the system will check the number of processes. If the limit is
not yet reached, a new service instance is created, which will handle
the request: Read the message, process it, return the information and
listen for the next request. If the limit of number of processes has
been reached, there will be NO new proces created and the request will
fail since no connection can be established.

On nodeA, both services S1 and S2 have a number of instances available
This node is actually of no interrest, since S3 has been disabled on that
machine. Normally, it would also be handling NodeC's requests if NodeB
would not respond fast enough....

On NodeB, service S1 has 12 instances running of service S1. If all of
these would be activated (meaning: is busy handling a request so will NOT
react on the request) a new service-instance would be created to handle
a newly incoming request. If any would be idle (not busy handling a
request, therefore listening for a new one) it would take it, so no new
service instance is required, hence not created.

Service S2 has 15 instances, being the limit sepcified, so new requests will
fail if all 15 would be busy handling requests.

Service S3 has just 4 instances and should allow new instances be started if
the application on NodeC would issue a request and all running instances are
busy handling requests. THIS FAILS. On NodeB, no logfile of service S3 is
created!

There are two possible locations for the problem: NodeB and NodeC.