I know that the faster the PCB is, the more issues you get. Above a certain frequency, inductors look like capacitors, capacitors look like inductors, and PCB-traces look like transmission lines with reflections and other such nonsense. Staying at a slower speed helps negate these issues.
Most application notes, be it from STMicro (for STM32) or Microchip, or really any other microcontroller manufacturerer, will have recommended hardware designs + their thought process fully documented.
Start there. Here's Microchip's ATMega328 hardware design notes: https://ww1.microchip.com/downloads/en/Appnotes/AN2519-AVR-M...
STM32F4: https://www.st.com/resource/en/application_note/an4488-getti...
---------------
Study up on the "reference designs". For ATMega328p, that's Arduino Uno. For more recent AVR chips (such as AVR DD), that's "AVR DD Curiosity Nano". (See schematics here: https://www.microchip.com/en-us/development-tool/EV72Y42A)
I made a sensor board the other day (I'm just printing the case for it now), and it was very enjoyable, and even came assembled for $1.7 per board:
https://gitlab.com/stavros/sensor-board
Feel free to email me if you have any questions or just want to chat.
Also, I don't think I've ever wanted something in my life more than this badge thing.
That's fair.
Lets put it this way: if your circuit works on a breadboard, you don't need to know anything about PCB design. The PCB will pretty much always be better than the breadboard.
Things get troublesome as you enter mixed-signal (analog + digital), or high-frequency.
CR2032 only has 235mA-hrs of life down to 2V. x2 and that's only 470mA-hrs and 2V probably browns-out the circuit.
Over 10 Years, that's 5uA of power usage on the average.
----------
Note that an aluminum capacitor has ~15 uA of leakage current, and Tantalum is ~1uA of leakage current. So Aluminum caps are already disqualified, and Tantalum capacitor leakage-current already uses 20% of your power budget. Given the "Burstiness" of this workload, I know that these capacitors need to exist somewhere.
You probably can get a year out of this in practice. To get better than that, you'll need to spend an incredible amount of energy on finding every 1uA "leak" and plugging the leak.
And its crazy how many things leak 1uA. Not only capacitors leak 1uA, but so do MOSFETs, diodes (reverse bias currents, especially in schottky diodes)... diode-protected MOSFETs (oh no, twice the leakage!).