Cache status not correctly updated when several CBs are working in parallel

[fgalan]

Let's consider two CBs using the same DB (typical scenario in HA configurations):

1. One active csub in DB
2. Start CB1
3. Start CB2
4. Update entity in CB1 -> notif is sent (OK)
5. Update entity in CB2 -> notif is sent (OK)
6. csub updated to inactive in CB1
7. Wait for a csub cache cycle
8. Update entity in CB1 -> notif is not sent (OK)
9. Update entity in CB2 -> notif is sent (NOK)

Additional info:

• If we swap the role of CB1 and CB2 in the above sequence the problem is the same (but in this case CB1 is the one which fails, sending a notification that must not be sent)
• If has been detected both in 3.5.1 and master
• The problem doesn't occur if -noCache is used

[fgalan]

Moreover if we continue the sequence:

10. csub updated to inactive in CB2
11. Wait for a csub cache cycle
12. Update entity in CB2 -> notif is not sent [i.e. CB2 is back to the right behaviour]
13. Update entity in CB1 -> notif is sent (NOK) [i.e. CB1 previously working right now works wrong]

[fgalan]

PR #4070