Okay, I figured out what happens 🙂. First I would like to correct myself: the measured (idle) voltage of 1.15V was actually 1.115V. This is good.
When powered down, I double-checked all regulators again and was surprised that even the SY8113 is still on, but showed a bit more than 1.3V! For some reason I must have made a mistake when measuring before as this is consistent now. Even when booting, the SY8113 shows 1.3V initially and then drops to 1.115V after a while with occasionally switching between 1.3V and 1.115V later on depending on the load.
So, I did take another more detailed look at the schematics and indeed Sinovoip implemented a poor man's PMIC for the 1.1V CPUX power supply that switches between these two voltages depending on the load and the CPU clock frequency. The signal is CPUX-SET. I guess this is known as the Linux kernel seems to switch between these two voltages.
Now, I'm still interested in low power consumption after poweroff. So far, the board uses more power after executing poweroff compared to running idle on Linux (~0.7-0.8W vs. ~0.5W). The reason is that all regulators keep being enabled plus the SY8113 remains to be configured for 1.3V.
I'm aware of the AR100 (CPUS) and also the crust firmware on GitHub. Obviously, the hardware implements that all but one (the 3.3V) regulator can be disabled, but so far that doesn't seem to happen. How can I help here? Should I look at and modify the crust firmware, i.e. add waiting to receive a "poweroff" trigger message from the kernel and set the appropriate GPIOs accordingly? The schematic tells me that CPUS is powered by RTC_VIO (via VDD_RTC via VCC-3V3). VCC-3V3 is the only regulator which can't be turned off. So, that's in principle okay as CPUS can still monitor POWER-KEY and restart the whole system.
Best,
Stephan