Re: Why does this need two Rovers?
I'd say the reasoning behind this methodology goes something like this:
The first stage, sending out a sample-collecting rover, is a bog-standard Mars mission, something we have experience with now and can carry out with a high degree of confidence. This rover, like Curiosity, would have analytical instruments on board so scientists could examine the samples in situ, so any "inert chunks of sandstone" can be tossed back and anything that looks like it might harbour signs of life or otherwise be of interest added to the cache for return.
But this process might take years before turning up something worth sending back, something that the instruments on board can't resolve - whatever that might be. If we sent an ascent stage with that first rover, that ascent module is going to be sitting on the Martian surface under durance of a hostile environment possibly for years. That's a lot of time for things to go wrong - dust clogging up engines, for example, or getting buried in a sandstorm, or blown over. Also, it's plenty of time for the volatile fuel the ascent stage needs for liftoff to leak and evaporate into the thin Martian atmosphere while it's sitting there.
Doing it this way also allows us to keep an open-ended schedule on the collecting mission. If the rover finds nothing of interest for 5 years, it's not a problem - we just keep looking until we find something worth sending back or the rover fails, whichever comes first. If it turns out to be the latter case, it's still not a wasted mission, because we were still able to do some good science with the rover while it was there.
If the rover has been able to collect some ambiguous samples that merit closer study, sending the ascent stage then means that the ascent vehicle is nice and fresh on arrival, and can make use of updated technology developed since the first rover launch. Which means much better odds for a successful lift-off and retrieval, than with a vehicle using older technology that has been sitting out collecting Martian dust and evaporating its hydrazine for X years.
The reason for a second pickup rover then becomes obvious: the ascent vehicle might have to land kilometres away from the collection rover. Perhaps the rover is in a crater or amongst large rocks which make a clean landing difficult or impossible. Or the rover has dumped a sample cache or three somewhere along its route in order to save power by not having to lug around a bloody great box of rocks on its back.
So having a second vehicle trundle out to the collection rover, or to wherever it's dropped its sample caches, makes sense in that regard. It also allows the second rover to be optimised as a taxi rather than a mobile laboratory. Current "laboratory" rovers have a top speed measured in centimetres per minute, to reduce the risk of tipping over or becoming trapped, and to ensure they don't miss anything of interest. A "taxi" rover could have its path pre-mapped by the first rover to avoid any obstacles and allow the taxi rover to run at a higher speed. This would allow it to perform its specific function much more effectively, since its only task is to pick up a box of rocks and return to the ascent stage.
So on consideration, while this mission might seem unnecessarily complex at first glance, it shows that the engineers and scientists involved have considered these issues and come up with quite an elegant solution.