Now that the MACs are random does that really solve the problem? A probe request sends out the real MAC of the AP it's looking for as well as the AP ESSID. By using anyone of many translators you can get a map of each ESSID with GPS co-ordinates. While many people will be probing for Starbucks and McDonalds they will always have a unique probe for their own home AP. So now there is no "neat" way of using the MAC as a primary key you can still infer the user by the AP least in common with anyone else, i.e. which probes are NOT McDonalds et al.
So if my home AP ESSID is Einstein, MAC=deadbeef every time I enter a store my home AP MAC is still being recorded as well as the relative movement throughout the store. As well inter-relational data could be inferred by other AP MAC addresses if I visit a friend or family member it's likely that probe will connect us.
TL;DR
Relations are based on unique data just because some of the data is 'scrambled' it's reliance on static data is it's weakness.
So if my home AP ESSID is Einstein, MAC=deadbeef every time I enter a store my home AP MAC is still being recorded as well as the relative movement throughout the store. As well inter-relational data could be inferred by other AP MAC addresses if I visit a friend or family member it's likely that probe will connect us.
TL;DR Relations are based on unique data just because some of the data is 'scrambled' it's reliance on static data is it's weakness.