Unnecessary Blocking in UncaughtExceptionHandlerIntegration When RateLimiter Is Active

[kkrawczy]

Integration

sentry

Java Version

19

Version

7.16.0

Steps to Reproduce

1. Configure your DSN with a low limit, 1.
2. Run a script that triggers many uncaughtException and observe how slow the handling gets. This line in the console can confirm slowdown: Timed out waiting to flush event to disk before crashing. Event: %s
   Simple script:

public class SentryTest {
    public static void main(String[] args) throws InterruptedException {
        Sentry.init(options -> {
            options.setDsn("XYZ");
            options.setEnvironment("XYZ");
            options.setDebug(true);
        });
        for(int i = 0; i < 100; i++) {
            System.out.println("Iteration: " + i);
            Thread workerThread = new Thread(() -> {
                Thread thread = Thread.currentThread();
                thread.getUncaughtExceptionHandler().uncaughtException(thread, new Exception("This is a test exception"));
            });

            long startTime = System.nanoTime();
            workerThread.start();
            workerThread.join();

            long durationInMillis = (System.nanoTime() - startTime) / 1_000_000;

            if (durationInMillis > 1000) {
                throw new IllegalStateException("Sentry blocked the thread");
            }
        }
    }
}

Expected Result

Don't block the thread, which could potentially slow down different parts of the user app.

Most probably, the latch used by BlockingFlushHint should be properly released in RateLimiter like you already do for other hints (markHintWhenSendingFailed)

Actual Result

The thread blocks for 15 seconds, blocking other parts of the user application.

The workaround could be to reduce flushTimeoutMillis.

[adinauer]

Thanks for opening this issue @kkrawczy, we'll look into it and update here once we did, can't say when we'll get to it yet.