NOTE
First run of the extended smoke (2026-11-09), before the two engine fixes, with an earlier firmware revision (one-byte RXNEIE probe):
spiN ... pend=0 FAILon all three: the classic STM32 SPI raised RXNEIE only for a gpio-net master, never for a slave onexternal_devices.i2c1 ... nack_er=0 FAIL: AF was set under ITERREN, but in scheduler mode the F1 I2C event chain never emitted the ER line (only the legacy walk did).dma1 spi1 frame=0x90160001 ... FAIL: the one-byte RXNEIE probe left the MAX31855 one byte into its frame — the engine does not deliver GPIO CS edges to STM32 SPI slaves, so the device re-frames every four bytes. The smoke now clocks whole frames only; CS routing is recorded as an open gap.