To improve performances on indexation, we should review our Solr data updates management.

While we want to index an object, we should add it to a queue during a synchronous event. But the queue itself is asynchronous.
The computation of data to update in Solr should call a provider only when it is its turn, and not before to avoid lost of time.

When we add an object to the queue (with his provider), if the object is already in the queue, we should put it at the end of the queue all its indexation updates and group it to avoid duplicated operations.