Requirement status cache not being properly replicated across nodes

The steps below were reproduced in a datacenter environment with two nodes and one load balancer.

The easiest way to reproduce this is being skipping the load balancer and access the nodes directly. 

1. requirement A is in OK status and has one test covering it with PASS status (test B)
2. open requirement A in both nodes (both are OK)
3. access the test in node 1 and execute it to FAIL
4. do not open the requirement in node 1 from now on
5. open requirement A in node 2, the status should be NOK 
6. go to node 1 and execute test B to PASS
7. open requirement A in node 2, the status is still NOK which is wrong