Having seen ‘Auto Recovery’ feature in previous post, I tried the following use case to observe the behavior as we have some BPEL processes that follow this pattern. An asynchronous BPEL process (BPELA) calls synchronous BPEL (BPELB) having mid dehydration point and both of these BPEL processes don’t have the fault handlers.
BPELA:
BPELB:
Flow Trace:
If we observe, the auto recovery is happening for Sync BPEL from dehydration point and also the entire flow. This is because of the fault is not caught either in caller or calle which was propagated to runtime. One should be careful about such scenario if we are relying on auto recovery. This can be avoided by using reply activity to send the fault from sync process back to caller.
Siva's Blog 
Hi there, just wanted to tell you, I loved this post. It was practical.
Keep on posting!