Large message reduction test gap?

[jdinan]

Missed #680 apparently because of a gap in testing coverage over large message reductions. Is this case covered or should we add a test that sweeps a range of message sizes? (same may be true for other collectives as well)