I can reproduce the problem, which did not happen with Wheezy. UEFI boot, fresh Debian 8 amd64, RAID 1 on two GPT disks.
Another person experienced it too on disks with legacy MBR/MSDOS partition scheme, so I do not think it is related to GPT. Note that this does not happen when the missing member was previously recorded as faulty. The problem seems to be caused by the incremental assembly performed by udev in /lib/udev/rules.d/64-md-raid-assembly.rules. When a member is missing, it leaves the array in an "inactive" state. The classic assembly performed in the mdadm initramfs script (/usr/share/initramfs-tools/scripts/local-top/mdadm) does not activate it, probably because the array already exists. Should the mdadm script force activation of required degraded arrays with mdadm --run ? -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org