Request to have app profile level settings to limit number of tasks executed in a persistent service instance

In a large cluster shared by 100's of applications from different LOBs, quite often, the a bad application workload might crash or disable servers by:
1) over consumption of system memory
2) filling up the root file systems with core files
3) etc
In the situation of over consumption of system memory case, even with the cgroup memory limit feature over symphony workload is set. We still see cases of system memory over consumption and crashes the server rapidly. The outcome of this is that a bad behaving application workload bring down servers along with the workloads from other applications (especially over long running tasks).
Per our observation, such bad application has the following characteristics:
1) The service instance persists to avoid the latency of starting the service instances
2) The run-time of its tasks is very short (in seconds if not in minutes)
3) If there is memory leak and/or slowness of freeing unused memory, it will resource a large accumulative memory foot print on the system.
4) The bad behavior will also leave no/less memory for other application workloads on the same server/system.
In order to avoid a bad application going wild, a new demand arise to have ability at the application configuration level to stop abusing the server by limiting the maximum number of tasks executed in a persistent service instance. So that, the application could release the memory faster and have the shared server more reliable to other good applications.