Thanks for the suggestion, but the problem still exists.
I did some further testing with different notebooks and extremely good signal strength:
Local websites for example the APs webinterface *sometimes* get stuck loading for about 10 seconds (refreshing the site ~ 20 times = 1 hang), despite excellent snr (>40, signal ~ -50, noise ~ -95).
ping delays are inconsistent: between 3-8ms.
same host with an old 802.11g AP, connected to the same switch: <1ms
This all happens with pure N (5Ghz), N/a, a, short or long guard interval, different rts treshold settings I tried, different channels -though site surveys have shown that there's nothing else on that channels, there's even no other AP in the whole 5GHz band, different notebooks (but all use intel cards, namely 5100 and 5300, all latest drivers directly from intel's website + any power savings turned off (psp -> cam, et.c.))
Still, single retry frames count upwards at a rate of *at least* 1/second at night with only me connected (idle, except ping).
Firmware is latest (5.4.1)
All this also happens with a MSM410, same firmware.
I also moved them from a small cisco poe gigabit swith to hp power injectors + procurve 1810g-24 switches. Didn't change anything.
I'm really desperate now since these issues won't resolve despite trying for quite a long time and an upgrade from 5.4.0 to 5.4.1.
Of course it's too early to draw any conclusions, but I already regret buying those units.
Reading the release notes known issues:
"Under a very heavy wireless load, wireless client traffic can become very slow after
several large file transfers. Terminating and reassociating client stations does not
resolve the issue. The only solution is to restart the AP."
Even though my problems aren't related as I often rebooted the AP before testing (because of that known issue), this sounds really bad. I reboot the APs every night just to be a bit more save.
Does noone else have issues like that?