A full Confluence reindex is very occasionally required. The reindex is multithreaded to make it faster, however this sometimes leads to problems of starving requests of CPU and RAM. See CONF-19674 and http://confluence.atlassian.com/display/CONFKB/Confluence+Rebuild+Indexing+task+fails+to+complete.

The default is currently 1.5 * the number of cores available, which seems like it is still too many on a busy instance, so I think it should default to a lower number.

More importantly, while Confluence could try to make better guesses about how many threads should be used for a given index, it's very likely that the best guess will be made by the System Administrator triggering the re-index. In particular they will have a good idea of whether they're running the re-index at a time when minimal external load is expected and they want it to be as fast as possible, or whether they're running it during serious load and they don't care how long it takes.

One option is to give them a choice between "fast" and "low-impact", but I suspect that abstraction will be unhelpful and unnecessary for the target-user.