Why would light peak be implemented on usb3 it has no use for all those pins or all that copper. Remember copper is starting to get expensive. The point is have an optical spec ready to to replace longer distance applications as the price rises. Also to as little copper as possible for short length cable devices that need power or fallback data if the cost is to high.
Copper is getting expensive? The whole reason they are putting "Light" Peak (which is laughable given it will be on copper for quite some time) on copper is because it is too expensive to use sufficiently fast optical fiber transmitters and receivers (the cable itself isn't really any more expensive than copper) in order to operate at Light Peak speeds. In any case, Light Peak would want to implement compatibility with USB3, not the other way around. The faster (higher bandwidth) standard needs to incorporate the slower one (the other way around makes no sense since it would defeat the point).
It would make more sense to backbone it on usb2 better still for oem like apple and Sony who want to make small device for the higher end market who'll pay for it microUSB. Maybe a more robust multiple bus cable for long runs like Ethernet copper replacement after all it's not often you run a single Ethernet cable to a location without wishing pretty soon you had another one or four there.
Ethernet cable is dirt cheap in any length run that a consumer would use (a quick glance on Google is showing me 1000 foot of bulk cable for a mere $82), so your arguments fall flat there. The latter part sounds ridiculous as well since one would not run multiple Ethernet cables to a distant location. One would use ONE cable and connect a fast switch (Gigabit or faster) at the other end of the cable and then use short runs of cable. After all, the switch has to be at one end or the other anyway so why would anyone want to run multiple long cables when they can run one short one and put the switch on the far end? You can even have multiple switches per room and have any number of Ethernet connections. Short of using an entirely separate bus (probably not needed in a consumer application or house), one wire to a distant area should be sufficient or depending on the house layout, perhaps one wire to a junction switch near the center of the house and branch out from there (especially for in-wall applications) with perhaps secondary switches for any given room that might need multiple devices.
In any case, the type of connector is largely moot. It would make way more sense for Light Peak to use existing CAT6 cable for copper runs than inventing a new connector since the latter would only add huge price increases to existing installations (not to mention any new cable costs a fortune until it's made in generic form since companies always gouge with custom connectors). In fact, a combination Ethernet/USB connection set with a single Light Peak bus makes sense for the motherboard. Consumers will need multiple connectors and types anyway, especially until devices support a unified connector.
Otherwise, you'd have a myriad of adapter/dongles and that would be one freaking headache for the consumer. And let me tell you that if you think consumers don't mind using/carrying a crap load of dongles instead of just having all the connections there and ready, you're kidding yourself. It will fail HARD. New standards need to benefit the consumer, not make their lives miserable (which is what requiring a load of dongles/adapters does; just look at the whining that goes on with FW800 only notebooks in order to connect a FW400 device. It sucks to have to use an adapter. A smart standard thinks ahead about future connectors so the older cables will just plug right in (e.g. USB 2 and 1 (and now 3 as well) versus say FW800 versus FW400. Look which standard sells and which does not, despite the technical superiority of FW800 to USB 2.).