I know the path you are wandering. I have been considering to do it myself. But the fact is that the in-wall boxes on your side of the pond are a fair bit bigger than here in the EU. It would be impossible for me to fit the button assembly (which would stick about 12mm into the box) with some form of battery and a 2-channel third party switching device with all the wiring inside the 50mm deep and 72mm across box.
That's why I made my own. See the pictures attached.
Right now I do not have an example of the complete assembly as it is in a full-on test.
This board of 50x50mm is the biggest I can fit inside the wall box. The idea is to first place the board inside the box and then route the wires over the pcb and connect the button/brain board last. The board is attached to the momentary switches so that closes the box neatly.
Another consideration for me is a financial one. My current actuators are de CoCo/Intertechno 433MHz ones without any delivery confirmation. There are days that messages arrive only intermittently, and for a whole house automation, that is simply not good enough. If I have to choose a new framework, other than MySensors, it would have to be Z-Wave in my opinion, as the 2.4GHz in my apartment building is very crowded (160 units). That means I have to invest in a gateway and actuators which cost a minimum of €60 each. These units can also be used as button senders, but lack the option of a signalling device (I have a RGB led option that lights between the button and the frame (I so need to make a little video of this!))
One of the top priorities in my design spec is that even with the controller down, my better half is able to control all the lighting in the house, as there are situations that I am away from home for a few days (also another reason not to go with 2.4GHz devices). All reasons to make my own design which is adaptable to a lot of situations.
And now I go and make an end to this very lengthy post. My humble apologies if I fail to make sense by jumping from topic to topic...