Thanks again,
Firstly the answers to your questions...
Processes require persistence. We want to ensure that we lose no data on failure and plan to make them restartable
There are no human tasks.
Each service task lasts up to a few seconds. We use timers a lot to poll for asynch responses so total process execution time may run to many minutes.
I will pursue the async handler approach. Just for my satisfaction, though, what are the blockers to my current approach. It is generally much easier to write sync code and let the workflow engine m-thread.
I appreciate your views.
Cheers,
Richard