@RHL suggests (via hipchat moments ago) "Deploying the SQL backend is the real answer, but I need to do some more work and testing.  Lots of tags seem to be a problem but I haven't profiled to know why.  I'd try removing tags, and bugging me to work on the SQL stuff."

The tag removal technique as tried a few weeks ago and didn't last very long. We are ready for the longer term solution.

Removing tags is not enough (although we've added a lot). Removing versions is also necessary since the worst code (at https://github.com/RobertLuptonTheGood/eups/blob/master/python/eups/db/Database.py#L678) looks at all tags and all versions.

Is there a shorter-term solution that avoids the above code?

I didn't write that code and I'm not surprised that it's a disaster. Thanks for profiling, that's where I would have started.

A SQL backend is almost certainly needed anyway, both for performance and to solve the locking issues properly, but maybe not urgently.

Kian-Tat Lim, Robert Lupton is this still an issue?  If not (or if we are not planning to fix it) I will set to Won't Fix 

This is still an issue. I don't believe it will be going away without a relatively large investment of RHL time, which I think must be quite unlikely.

Robert Lupton informs me that there are workarounds in place and that this is not a significant blocker for anyone. This will not be fixed without some significant time investment and is not a priority at this time.