During work with jBPM process with few users, at some point the system starts to make millions of write operations to table processinstanceinfo which results the table space where the LOB PROCESSINSTANCEBYTEARRAY is stored to grow exponentially causing the server to crash due to lack of storage.

I am attaching the AWR report for that time frame ( 2 hours), where you can see the results in tab:
"Segments by physical writes"

I am also attaching the records from processinstanceinfo for the period of 2 hours before crash.