Can't assemble /dev/md0 after update to Turris 4 on Omnia

I tried updating my Turris Omnia NAS to TurrisOS 4.0.3. Everything seems to works fine, except for assembling the RAID, where mdadm fails with a segfault:

root@asgard:~# mdadm --examine --scan -v
ARRAY /dev/md/0  level=raid0 metadata=1.2 num-devices=2 UUID=551d792c:123e4485:509cf30f:796b845c name=turris:0
   devices=/dev/sdb1,/dev/sda1
root@asgard:~# mdadm --assemble /dev/md0 -v
mdadm: looking for devices for /dev/md0
mdadm: /dev/sdb is not one of /dev/sda1,/dev/sdb1
mdadm: /dev/sda is not one of /dev/sda1,/dev/sdb1
mdadm: /dev/mmcblk0boot0 is not one of /dev/sda1,/dev/sdb1
mdadm: /dev/mmcblk0boot1 is not one of /dev/sda1,/dev/sdb1
mdadm: /dev/mmcblk0rpmb is not one of /dev/sda1,/dev/sdb1
mdadm: /dev/mmcblk0p1 is not one of /dev/sda1,/dev/sdb1
mdadm: /dev/mmcblk0 is not one of /dev/sda1,/dev/sdb1
mdadm: /dev/mtdblock1 is not one of /dev/sda1,/dev/sdb1
mdadm: /dev/mtdblock0 is not one of /dev/sda1,/dev/sdb1
mdadm: /dev/sdb1 is identified as a member of /dev/md0, slot 1.
mdadm: /dev/sda1 is identified as a member of /dev/md0, slot 0.
Segmentation fault

If I roll back to the pre-4 snapshot, the array works.

Force-assembling as described at Restore RAID after migrating to TOS 4 seems to help, but the configuration won’t survive reboot even with kernel.randomize_va_space=0 in /etc/sysctl.conf as suggested in that thread.