I forgot to blacklist the "eepro100" module by default. This results in the old and deprecated "eepro100" driver being loaded on systems with the relevant hardware, together with the proper "e100" driver - and it is impossible to predict which driver grabs the hardware. Since the hardware works, and the bug existed on all previous CDs, this is not a blocker for r2052 being the first publicly available CD with the 6.3 book.
(and this deprecated driver even warns about the receiver lockup bug on ancient i586 box)
