ISSUE-10: Race condition potential in ApplicationManager concurrent operations

[sfloess]

Severity: Medium

Type: Concurrency / Thread Safety

Description:
While ApplicationManager uses fine-grained locking per application, there are potential race conditions in operations that span multiple applications or access shared state.

Location:

• platform-core/src/main/java/org/flossware/platform/core/ApplicationManager.java:970-986 (shutdown method)
• Various places where lifecycle listeners are notified

Problem:

1. The shutdown() method iterates over applications while other threads might be deploying/undeploying apps
2. The ListenerExecutor uses unbounded task submission without backpressure
3. Multiple lock acquisitions in sequence can deadlock if ordering is inconsistent across methods

Example race:
Thread 1: shutdown() iterates applications list
Thread 2: deploy() adds a new application
Result: Application might be deployed then immediately undeployed

Impact:
Medium - Can cause:

• Applications left in inconsistent state during shutdown
• Listener notifications lost
• Potential deadlocks if care isn't taken with lock ordering

Recommendation:

1. Use consistent lock ordering across all methods that acquire multiple locks
2. Use snapshot-based iteration (already done on line 974) consistently everywhere
3. Consider a RW-lock for the applications map during bulk operations
4. Document lock ordering to prevent deadlocks in future maintenance