Claim-time volume mount occasionally sees the device before it is a valid block device #400

A claim with a writable volume failed once at mount time during the #398 hardware round:

setup volume "hwrt": mount volume hwrt: exit code 32: mount: /volumes/hwrt: /dev/vdc is not a valid block device.

• Frequency: 1 of about 270 claims (cocoon v0.7.4, CH v54, rt:24.04 sha256:03e8bed49d4c, ext4 image, none pool, small). The next 120 claims on both builds had none.
• The claim-time attach and mount path (applyVolume → DiskAttach → MountVolume) did not change in pool: one rule per live volume invariant, whole-repo audit #398, and main showed no failure over a similar count, so this looks like an existing race: the device node appears in the guest before the block device is ready for mount.
• The claim failed cleanly (500, rollback released the hold); the next claim succeeded.

To do: reproduce with a tight claim/release loop, record guest dmesg around the attach, and decide whether waitForVolumeDevice should wait for a readable size (or retry mount once on this error) before mounting.