Each time message is opened from Mail Archive, it builds thread by searching related messages and organising them into tree. This searching/tree building process takes quadratic time and memory relative to how many related messages have been found. Currently we treat related messages as either

• Having reference between each other using 'References' email field
• Having similar subject to each other

The problem here is that there's no upper limit on number of related messages. If their number is about 20K, it takes several minutes for thread building process to finish. As this operation is Lucene and CPU heavy, it can bring Confluence down.

We should at least provide some way to limit the number of messages in thread in order to save application from OOMs and CPU spikes.

System variable

-Dconfluence.mail.archive.thread.limit=250