Potential performance problems with Workspace.copy(...) of content with multiple 'mix:versionable' nodes

BRMS is seeing performance problems with repeated operations involving Workspace.copy() of content with multiple 'mix:versionable' nodes, and subsequent operations (including checkout, checkin) involving the versionable nodes. Details (including charts) can be found in BRMS-502. Recent tests with various configurations have shown that the performance degradation is not seen when using the in-memory connector (with in-memory or file-based indexes), but is seen when using Oracle 10g, Oracle 11g, and MySQL 5.1. I'm creating this issue as a placeholder for investigating this issue and, if an issue is found, for the fix. I'm also speculating at this time that the problem is the JPA connector, tho that may change with further investigation.

(Originally, this was thought to be a problem with the JPA connector, since these connectors were being used with databases that had really high latency. However, I've replicated the problem with the In-Memory connector, though the problem is not nearly as obvious, noticeable, or impactful as when using the JPA connector. As such, I've changed the affected components from 'Connectors' to 'JCR', and re-titled the issue.)