Currently, when a new transaction is started and completed in runtimes to a given workflow instance, a set of events from the internal engine related to the transaction are mapped to external events to be published and consumed by the Data Index.

On the Data Index side some issues might happen when consuming events to a given process instance Id asynchronously and in parallel, since multiple threads may be accessing and updating/deleting/inserting the same registry at the DB level, using the same primary key.

There should be tackled mainly to areas:

• Concurrency
• Ordering of events to be indexed to avoid overriding or losing information.