AFAIK Quartz job are always executed on a single node that is picked by Quartz. The @NonConcurrent annotation only prevents Quartz from executing the same job concurrently on a particular node.
In other words, you cannot make Quartz execute a job on multiple nodes concurrently. It always picks a single node to execute the job on.
To realize what you describe, you may need multiple jobs (using the same job class and without associated triggers). Then you will need to implement some sort of an orchestrator job that would remotely connect, e.g. via JMX or RMI, to individual nodes and trigger the jobs manually.
You may want to check our product QuartzDesk (www.quartzdesk.com) that provides a web-service that provides a single-endpoint through which you can connect to individual Quartz scheduler instances and, for example, trigger jobs on them.