I have seen many Wifi Benchmarks for speed only and not for reliability.
I stopped doing wifi benchmarks for speed since i get very different values for the same board a few hours later,.
For real life projects reliability and availability is what matters when you depends solely in wifi network.
I would strongly suggest @tkaiser to include "reliability" in his tests, maybe a new thread.
As for example you can try iperf with -t 1000, -t 2000, -t 3000 and -t 10000 (if it allows?).
Try this with your board and see what happens...