Really the G.hn PL modulation seems more than adequate, but the behavior when conditions change is incredibly poor; if I had time, I'd find a "sufficiently noisy" appliance, some romex and run them all together. TCP and many video streaming sites are quite good at handling variable bandwidth, but the G.hn seems to turn that into variable latency, which is terrible. The people who design these are presumably not idiots, so I'd like to know why treating noise like congestion isn't sufficient...
The problem with P/L is that it can work, until suddenly it doesn't.