Here some notes I got to resolve such cases (verified at ovirt 4.2, should apply at 4.3 also). This should do.

# Recover VM from corrupted snapshot with disk in illegal state: (not clean solution as it leaves orphaned snapshots at the storage domains)

1. Login at engine DB
ssh to engine and run:

su - postgres -c 'scl enable rh-postgresql95 -- psql'
\c engine

2. Get the VM ID (from GUI or DB) and  find the broken snapshot in the snapshots table and delete it.

engine=# select snapshot_id,snapshot_type,status,description from snapshots where vm_id='0aff510b-b55a-4986-9d12-8407748fcd8b';                                         
2dd4c31d-7c57-4f87-b863-4fb3de1a5196 | REGULAR       | OK     | MyVM
 6a93187c-2b3f-409f-9831-3c11d36f3efb | ACTIVE        | OK     | Active VM
 818c46ff-4c32-496d-a4ca-12459e4ca917 | REGULAR       | OK     | Backup

4. Get the UUID of the affected VM disk (from GUI or DB) and find the image linked to the broken snapshot

select image_guid,parentid,imagestatus,vm_snapshot_id,volume_type,volume_format,active from images where image_group_id='0994f9c0-568e-4e4c-b0ea-0370de64d2a7';
 cf8707f2-bf1f-4827-8dc2-d7e6ffcc3d43 | 3f54c98e-07ca-4810-82d8-cbf3964c7ce5 |           1 | 2dd4c31d-7c57-4f87-b863-4fb3de1a5196 |           2 |             4 | f
 1e75898c-9790-4163-ad41-847cfe84db40 | cf8707f2-bf1f-4827-8dc2-d7e6ffcc3d43 |           4 | 818c46ff-4c32-496d-a4ca-12459e4ca917 |           2 |             4 | f
 604d84c3-8d5f-4bb6-a2b5-0aea79104e43 | 1e75898c-9790-4163-ad41-847cfe84db40 |           1 | 6a93187c-2b3f-409f-9831-3c11d36f3efb |           2 |             4 | t

5. Delete the broken snapshot from the snapshots table.
engine# delete from snapshots where snapshot_id='818c46ff-4c32-496d-a4ca-12459e4ca917';
DELETE 1

6. Delete the associated image to the broken snapshots.
engine=# delete from images where image_guid='1e75898c-9790-4163-ad41-847cfe84db40';
DELETE 1

At this time, the snapshot is no longer shown on the 'Snapshots' tab of the VM.

7. check the status of volumes in the data domain. At SPM run:

vdsm-tool -vvv dump-volume-chains 142bbde6-ef9d-4a52-b9da-2de533c1f1bd | grep ILLEGAL

8. In case no ILLLEGAL volumes are listed just start the VM.
In case you still get ILLEGAL volumes of the affected VM, then make them legal:

To get the volume ID startup the VM and observe vdsm log:  

tail -f /var/log/vdsm/vdsm.log |grep prepareImage

Example:
INFO (vm/5bf9a0bb) [vdsm.api] START prepareImage(sdUUID=u'bc0480e2-85fe-42a4-91ae-f733b23c801f', spUUID=u'75bf8f48-970f-42bc-8596-f8ab6efb2b63', imgUUID=u'870f7e85-d9b6-494a-9541-b419fb0e1b32', leafUUID=u'd7fa8c51-8cad-4695-b90a-a8d1dc146371', allowIllegal=False) from=internal, task_id=89770233-103d-47f3-acc1-45d2e96d9e91 (api:46)

Now, go to the SPM and run a command like this:

vdsm-client -a host0.localdomain.local Volume setLegality storagedomainID=bc0480e2-85fe-42a4-91ae-f733b23c801f storagepoolID=75bf8f48-970f-42bc-8596-f8ab6efb2b63 imageID=870f7e85-d9b6-494a-9541-b419fb0e1b32 volumeID=d7fa8c51-8cad-4695-b90a-a8d1dc146371 legality=LEGAL

In case you still get ILLEGAL image status, although no ILLEGAL images are reported from SPM and not chain issues are observed at image file, then you may just flag the image as ok in the DB:
Example: update images set imagestatus='1' where image_guid='snapshot id';