Several nRF91 DKs share a runner, so a rig whose CI_*_SERIAL_MODEM_SEGGER_SN names the wrong board is the hardest failure to spot: nothing errors. That board is programmed and captured while the host talks to a different one, so the log holds the boot that programming triggered and then nothing. AT#XLOG=1 is still acknowledged, and AT#XLOG? still reads back 1, because both travel over the host's CMUX pipe to the board actually attached, and the test passes because that board does the work. Two nRF54L15 rigs had their SEGGER SNs swapped this way, which is why they never logged past boot while the nRF54LM20 rig did.
Identities are therefore compared rather than assumed. AT+CGSN=1 gives the IMEI of the modem the host is attached to, Serial Modem prints its own IMEI during boot before suspending the console, and a mismatch logs an error naming both IMEIs and the variable to correct. When that fires, find the right serial number by matching the IMEI the host reports against the boot output each candidate DK produces.