Hi AG,
I like your style!! Progress already. I got onto some clone Arduino Pro-Mini versions that are a good size and also have the 4 and 5 exposed in a double strip across the bottom.
https://www.ebay.com.au/itm/Pro-Mini-atmega328-Replace-ATmega128-5V-16M-For-Arduino-Compatible-Nano-DA-/162507940764?hash=item25d63a639cThe only drawback is that I have to press the reset (at the right time) to get it to download, just after a compile. However they are compact, and with the extra pins quite versatile. There are many Pro-Mini boards on sale, but not all are like this one.
I haven't tried to create an I2C (or SPI) device, so I can't help you there. However I have hooked up my Temp transmitter board to the parent Arduino, and removed the 433MHz Tx and RX boards all together and it worked fine, very reliable using the Manchester code. I did that just to debug the Tx code without blasting the 'real' Rx in the shed with test signals.
The use of interrupts would be good. I only ended up using a 1 minute "report to RPi3" interval as my old BIOS weather station had all readings combined and sent them once every minute. My Python code was set up to use it as the time base, so I fashioned the later Oregon Scientific sensor software to replicate the same thing. My whole Arduino to RPi3 system is centered on the weather sensors so it does not matter if they dominate the situation (plus the weather data for most sensors is only graphed for 12 hours and so the accumulated errors of using the Arduino time base is fairly small over that 12 hours). Daily long term samples (eg rainfall) are triggered from the clock on the RPI3 which is synced with NNTP. My Arduino base station just reuses the last reading if a new one has not turned up in the last minute. eg anemometer reports every 14 seconds, so rarely (see below though) misses, however UV is 75 seconds cycle but still acceptable for my purposes.
SPI would be good, and does not insist on interrupts? I think I would prefer SPI, it does not appear as mysterious as I2C. RPI has good SPI facilities. You can choose ;-). I have used the IR receive interrupt circuits (I designed an advert killer that mutes the TV when the cable adverts come on for 2, 2:30,3:00,3:30 minutes by pressing the Red, Green,Yellow or Blue buttons respectively on the Sony remote. So it does both, Rx from the human+remote, then translates that to Tx to mute function. Press the coloured button and it mutes, but comes back on auto-magically :-) eg when still making coffee!!)
The flexible antenna is good. I suspect the rigid helical antennas inside the Oregon Scientific sensors are too rigid and in the extremes of the weather, crack the solder or PCB and become intermittent. Flexible is better I reckon. Though harder to sell to the masses, not quite so tidy!! I have fixed my UV sensor this way. My anemometer has recently been intermittent so I bought new Lithiums yesterday to replace the batteries. When I got them down from the roof and tested them, they were very highly charged, so I am suspecting another 17.3cm "whip" antenna coming up very shortly!!! (so to speak).
Cheers, Rob