Here is why Project Ara/Phonebloks will most likely to fail.
I Hope I'm wrong, but the way i see Modular phone, is that it creates more complexity and introduce it's own problems:
1. More bulky and less power efficient
The strength of mobile phones (compact and power efficient), is that they use SoC (system-on-a-chip), but with the modular phone the SoC module will not work (or works in a less efficient way), Also since different modules need to communicate using universal protocol (oppose to specialized interface on the motherboard), they will need 'smart control chip' of their own, adding more complexity and bulk and less power efficiency .
Also in the Desktop computers (where the modular concept already exist) we don't care that much about size, this is way we can afford the large expansion slots and bays, we learned from the 'ultraportable' laptops that the more lighter/tighter/smaller the device is , the less modular and expandable it will be , integrated components can always be assembled more efficiently and in a smaller space than modular counterparts.
2. Standards takes time to develop, and they keep changing
For a successful modular phone, you need to create a standards and protocols for the blocks/modules, like sizes, communication, power draw standards, ...etc , such standards, first need time to be developed and to be accepted in the industry (3-5 years maybe?), even then these standards will keep be updated ,and with the fast base mobile technology is developing, you will find soon that your phone motherboard/core is not supporting the new standards, defeating the very purpose of modular phones.
3. Weight distribution
Each module have it's own weight, and with the all theses different parts attached to a small device in different possibilities, you will properly end up with an unbalanced phone.
4. It will cost more
Since there will be a lot of variants , there will be less mass production of each part, for example it will cost more to produce 10 different type of displays (each at a rate of 1 million/month), than to produce one display at a rate of 10 million/month, also selling each module by it self will make it more expansive, since you got it from the retail channel which add an overhead cost. I would also imagine that U.S. carriers will not like the modular model and will not subsidized such devices, or they will try to control it in an ugly way.