[ https://issues.jboss.org/browse/TEIID-5449?page=com.atlassian.jira.plugin... ] Van Halbert edited comment on TEIID-5449 at 10/4/18 9:25 AM: ------------------------------------------------------------- In a production environment, once a cache has data, it isn't going to change without a migration. So if a VDB has a related cache table updated, it must be backwards compatible. Also, the caveat being, there can't be 2 entities with the same name. Therefore, 2 simultaneous connections would not fail. However, in a development environment, there's possibility for collisions when 2 people would be developing using the same entity name, regardless if they are trying to use the same cache or not. The collision could occur due to the wrong marshaller instance being used (last in wins). Issue: the marshaller registration uses a single instance, so the last one registered for the same entity name will be used. Could use MashallerProvider , but as indicated, reads are async, so the thread based model doesn't work. was (Author: van.halbert): In a production environment, once a cache has data, it isn't going to change without a migration. So if a VDB has a related cache table updated, it must be backwards compatible. Therefore, 2 simultaneous connections would not fail. However, in a development environment, there's possibility for collisions when 2 people would be using the same entity name. -- This message was sent by Atlassian Jira (v7.12.1#712002)