Process xx failed to create IPC queue __FCS_REGISTRY__ : Failed to allocate shared memory.

We are having issues recently on two of our servers. They both run with AMS 5.0.6 on a powerful Windows Server 2008 R2 with 40 GB of RAM and a lot of disk space. We split the overall charge of the servers by opening different instances on both servers. Recently we encountered a problem where the instance (core) cannot start on one of the two servers, having an error such as:

2014-12-17    09:23:45    81364    (e)2581279    Assert failed in tincan\util\TCShrMem.h line 332    - 2014-12-17    09:23:45    81364    (e)2801340    Process 81364 failed to create IPC queue __FCS_REGISTRY__ : Failed to allocate shared memory.    - 2014-12-17    09:23:45    81364    (e)2571122    Server aborted.    -

We then tried to transfer the opening of the instance on the second server but guess what, we start having the same error! So now both servers aren't starting any new core or even the admin service. We have to restart AMS and only then can we open new cores on the server. I think the cores left opened are still working but it is impossible to open new ones.

Also, when we stop the AMS service, all the cores stay opened and we have to kill them manually, which isn't the case normally.

We know it's really not an available memory problem since there is still about 30 GB of RAM available and a lot of space on disk too (about 60 GB which is for the system and services only, no data). What is weird about that is having the same problem on two different machines at the same time! Other than the shared data disk for stream source (about 6 TB left), they do not have anything in common.

Any help would be great on this. We had this very same problem twice now in the last month and it is impacting our clients directly since they cannot access their activity when that happens. The service is simply down and we are in trouble! We have to find a solution and fast! Thanks!