The consistency issue I was referring to related to the "timeout" argument - we currently have a patch applied to Fedora (and derivatives) that instead adds a "balanced=True" flag argument so you can pass "polling=False" to turn off the busy loop (at the risk of a long wake-up delay).

This means the current situation we have in Fedora (et al) is:

* Python 2 only code can pass "balanced=False" (but probably wouldn't want to due to the potential increase in wake-up latency)
* Python 3 only code can specify a timeout directly
* Single source Python 2/3 code can't do either (since the API signatures are different)

So downstream I'll be advocating in favour of Flavio's pthread patch regardless of what we do upstream - that way hybrid Python 2/3 code that's specific to the Fedora derived ecosystem (and we have a lot of that in the operating system layer) can eventually be updated to use threading timeouts while remaining consistent between Fedora (which is switching the system Python to Python 3) and RHEL/CentOS 7 (which will continue to use Python 2.7).

For upstream, I think it would be desirable to backport the timeout argument and simply have it throw RuntimeError if used with any of the backends other than thread_nt and thread_pthread. That way single source Python 2/3 code gains access to the timeout option, while code relying on one of the no-longer-available-in-Python-3 threading backends doesn't incur any stability risks.

However, I don't feel strongly enough about that to argue in favour of it against opposition - I'm happy enough to confine my argument to changing the downstream Fedora/RHEL/CentOS patch.