Problems with synchronization between provisionmonitors when a cybernode dies

[prubach]

Steps to reproduce:

1. Clone 5.0-M4 as of 09/19/2014.
2. Build distribution using: mvn -P release -DskipTests=true clean install
3. Copy rio-5.0-M4.zip and unzip on 3 machines:

Lynx: Core i5 2.5GHz, 8 GB RAM, Ubuntu 14.04.1 amd64, Oracle Java 8 64 bit (build 1.8.0_20-b26) (IP: 192.168.0.5) Esprimo: Core i5 1.5GHz, 4 GB RAM, Windows 8.1 64bit, Oracle Java 8 64 bit (build 1.8.0_20-b26) (IP: 192.168.0.10) Yoga: Core2 2.1GHz, 4 GB RAM, Ubuntu 14.04.1 amd64, Oracle Java 8 64 bit (build 1.8.0_20-b26) (IP: 192.168.0.16) Clocks are synchronized.

1. Point RIO_HOME to the installation directory
2. Compile calculator on every machine: unzip examples.zip && cd examples && mvn clean install BTW: Calculator tests fail on Yoga (Windows) (stacktrace in the attached files).
3. Start RIO on all machines (startall)
4. Start RIO UI on Lynx
5. Deploy Calculator using rio-ui (OAR)
6. After services come up, kill Lynx (@02:41:17) (Add and Divide were started on this cybernode)
7. Wait 10 minutes for Add and Divide to come back on one of the other machines but no success...