So far I've not heard "this can't be done." All I've heard is "I don't know how this might be done." Huge difference there. I also hear a fair bit of "this shouldn't be done", which I find incredibly disappointing. Here's a hypothetical: we already know how to do noisy, shared-media device interconnects. It's called 802.11. Look at the size of wifi adapters: you can get a usable signal that works over meters from som…
Modern phones are marvels of stripping away everything not completely necessary, with the result that they're the size they are. Adding modularity means adding enclosures around the modules, which adds size and weight - and for something as small as a phone, practical physical limits of enclosures add up to a non-trivial increase in size. Separability between two components means adding shielding, casing, air gaps, etc for each module, to just (say) have a removable battery means at least 2mm greater size - not trivial when the whole device is less than 10mm thick.
Should you accept that additional volume & mass (and cost) required to implement any degree of modularity, you're faced with the tradeoff: would you rather have significantly greater performance instead? You could apply that aforementioned 2mm to 50% more battery, or quadrupling storage, or etc.; which is more desirable, a replaceable battery or a longer lasting one? Most customers would prefer more performance.
To your thoughtful point of solving the problem instead of complaining at it, consider that the phones are reaching the limits of physical and computational need/usability: with continued miniaturization of all components, we can free up enough space within the device to apply it to lower-priority use such as modularity and optical interconnections.