Cinder many volume's snapshots deleting issue

Detailed bug description:

Usually, deleting about 50 volume's snapshots took about 3 minutes. But sometimes it continues more than 20 minutes. Some of snapshots are in deleting status even after this waiting.

Steps to reproduce:

1. Deploy MOS 9.0 with 1 controller, 2 computes, 1 cinder nodes
2. Create a volume
3. Create 70 snapshots for it. Wait for all snapshots to became in available status
4. Launch deleting all of them
5. Launch creation of 50 snapshot
6. Wait for all old snapshots (from step 3) to be deleted
7. Wait for all new snapshots to become in available status
8. Launch deleting all snapshots of volume
9. Wait until volume has not snapshots

Expected results:

All steps passed.

Actual result:

in step 9 some snapshots are in deleting status even after 20 minutes of waiting.

Reproducibility:

Sometimes

Description of the environment:
- Versions of components: MOS 9.0 #481, 477, 473
- Network model: VLAN