I'm seeing a lot of variation in times to just to a query against Bug.fti - up to five seconds. I think this is the root cause, not a side effect of the join. I don't see this variation querying bugtask. Perhaps it is because the Bug.fti values are generally much bigger, indexing the bug title and description, or perhaps the table or indexes are bloated. I'll check for bloat now.
I'm seeing a lot of variation in times to just to a query against Bug.fti - up to five seconds. I think this is the root cause, not a side effect of the join. I don't see this variation querying bugtask. Perhaps it is because the Bug.fti values are generally much bigger, indexing the bug title and description, or perhaps the table or indexes are bloated. I'll check for bloat now.