guest stuck at UEFI shell pre-boot doesn't migrate properly

When booting a VM, it's possible to get stuck at the UEFI shell prior to booting:

UEFI Interactive Shell v2.2
EDK II
UEFI v2.70 (EDK II, 0x00010000)
Mapping table
      FS0: Alias(s):HD0b:;BLK1:
          PciRoot(0x0)/Pci(0x11,0x0)/HD(1,GPT,E31ABB4E-EA6F-234A-C9F2-BEF14E665DA1,0x100,0x80000)
     BLK0: Alias(s):
          PciRoot(0x0)/Pci(0x11,0x0)
     BLK2: Alias(s):
          PciRoot(0x0)/Pci(0x11,0x0)/HD(2,GPT,05C06F5E-C665-1142-D74E-903AE9F80009,0x80100,0x363EDF)
     BLK3: Alias(s):
          PciRoot(0x0)/Pci(0x11,0x0)/HD(9,GPT,35F8DA03-37AF-D7C0-E49F-F3B61CB582AD,0x3E3FDF,0x4000)



Press ESC in 1 seconds to skip startup.nsh or any other key to continue.
Shell> 

(Note: It's usually possible to workaround this by typing exit, then selecting the the boot disk from the Boot Manager item on the UI.)

If a VM is migrated in this state, I've found that the serial console hangs at the destination, despite it working on the source. Rebooting the VM is sufficient to get the serial console working and get back to the UEFI shell.

I didn't debug this at all, and while this is not a "happy path" for migration, it would be pretty frustrating to not be able to use this console in such a situation, so I'm noting it as an issue for now.

Start by reproducing migration while the VM is at the UEFI shell, comparing serial-console behavior on the source and destination. Trace the migration path for this pre-boot state and the serial-console path; done means the console remains usable at the destination without requiring a reboot.