Right now SpiderMonkey is used in CouchDB, and I wonder if the CouchDB
community would be able to take advantage of TraceMonkey speed
improvements.
I think that's a question best asked on the CouchDB-dev mailing list. There
are tickets in JIRA for supporting spidermonkey 1.8. I don't know about
tracemonkey.
My experience suggests that JSON serialisation and deserialisation is the
biggest overhead for CouchDB view generation.