Even worse, an attacker can broadcast a spoofed deauth packet so that it looks like it's coming from your router. Most consumer wifi equipment will deauth and then immediately reconnect, allowing the attacker to capture the '4-way handshake' for offline cracking. For some reason I thought AES-256 was somehow part of WPA2, but not 100% sure.
It's also interesting how much information your personal wireless devices give out. If you use a program like airodump, you'll usually see at least one or two client devices with a long ESSID probe list from various restaurants, coffee shops etc.
Incidentally, this is how I came to realise how google use wifi to 'improve location accuracy' (although technically I think it's resolution time): their mapping vans would have catalogued router mac addresses/BSSIDs and bound them to locations. So google know that if someone's phone can 'see' a certain router, they must be within 100m or so of <location>. So a GPS fix can happen much quicker...
As for how practical it is to crack a WPA2 passphrase: by far the best protection is a totally random password (one that wasn't supplied by the manufacturer). Once you get up to a 10-12 character password (that's totally random), cracking it will take an infeasibly long time. If you're interested, you should check out some of the conversation of the hashcat forums. Some of the folks there use a number of very sophisticated ways of 'shrinking the search space', like statistical analysis of big password DB dumps. It turns out most people use passwords like 'Camero87' or 'Mystreetname1987'...
Oh, and disable WPS.