Filesystem access blocks indefinitely after failure of all legs in an lvm mirrored logical volume in RHEL 6

Issue

• After all legs/logs in a mirrored logical volume fail, the file system never aborts or remounts read-only. All oustanding I/O to the device is queued indefinitely and never returns.
• In-progress I/O to a mirrored logical volume never returns after all devices fail, when the repair operation is unsuccessful:

Dec 21 11:29:18 hostname lvm[8665]: Volume group "test" not found
Dec 21 11:29:18 hostname lvm[8665]: Repair of mirrored device test-lv1 failed.
Dec 21 11:29:18 hostname lvm[8665]: Failed to remove faulty devices in test-lv1.

• In our environment we found that when a cluster node loses access to disk devices, most times, fs resources hangs when trying to stop while executing sync.