The test boards arrived and Frank soldered it. Besides a small bug with a missing trace for the USB power supply (GND was not connected to the circuit GND) it worked. Test setup:
Flashing the ATMega was a bit tricky as well. Maybe because of the 1k series resistors to use the programming pins for the programmer and for the SPI FRAM in parallel without jumpers, it didn't work at the default speed. But by editing the file /home/username/bin/arduino-1.8.9/hardware/arduino/avr/programmers.txt and adding the "-B 10" command line option to the usbasp programming line, the programming speed was reduced and it worked. And another trick: if it is a fresh ATMega, you have to program the fuses the right way. This can be done with avrdude, the details are on the Github page.
For testing, there is a debug output jumper. If closed, it controls the MIDI output, but if open, it can be connected to a serial port receiver and you can see debug messages in the Arduino serial port monitor.
The planned LEDs were a bit dim as well, but was fine with brighter ones and the 7-segment LEDs and LED brightness could be matched with the right resistors. A full test of all hardware components is in the Github repository. Now the big board is in production.