fix(runtime): preserve exact callback ownership

[ReubenBond]

Depends on #10062

The visible diff includes the base PR until #10062 merges.

This follow-up makes callback ownership exact:

• Removes callbacks conditionally using both the correlation ID and expected CallbackData instance.
• Moves callback-owned cleanup through ICallbackDataTarget so cancellation, timeout, silo failure, and shutdown unregister only their own callback.
• Preserves replacement callbacks when stale lifecycle completion races with correlation ID reuse, providing ABA safety.
• Adds focused lifecycle coverage for response races, cancellation, timeout, shutdown, duplicate registration, and exact striped removal.

Microsoft Reviewers: Open in CodeFlow

[Copilot]

Copilot review overview

Review tier: Lite
Findings: 1

New issues introduced by this change (1)

Severity	Finding
	src/​Orleans.Core/​Messaging/​StripedCallbackDictionary.cs — TryRemove(id, expected) uses ReferenceEquals(value, expected), which will always return false…

What changed in this PR

This PR refines Orleans runtime callback lifecycle handling to make callback ownership exact (ABA-safe) when callbacks are removed due to cancellation, timeouts, shutdown, or silo failure. It builds on the striped callback tracking work from #10062 by ensuring that cleanup paths unregister only the specific CallbackData instance they own.

Changes:

• Introduces ICallbackDataTarget and routes callback-owned cleanup through it so unregister operations are instance-exact.
• Updates InsideRuntimeClient and OutsideRuntimeClient callback registries to enforce exact/unambiguous unregister semantics and detect duplicate registrations.
• Adds focused unit tests for striped dictionary behavior and callback lifecycle races (response vs timeout/cancellation/shutdown, replacement safety, duplicate registration).

File	Description
test/​Orleans.Runtime.Tests/​StripedCallbackDictionaryTests.cs	Adds coverage for stripe distribution, snapshot enumeration, concurrency, and exact-removal semantics.
test/​Orleans.Runtime.Tests/​CallbackDataTests.cs	Adds targeted lifecycle race/ABA tests and updates test harness to use ICallbackDataTarget.
src/​Orleans.Runtime/​Core/​InsideRuntimeClient.cs	Switches callback tracking to StripedCallbackDictionary<CallbackData> and implements instance-exact unregister via ICallbackDataTarget.
src/​Orleans.Core/​Runtime/​SharedCallbackData.cs	Removes the unregister delegate from shared state (unregister now flows through a target).
src/​Orleans.Core/​Runtime/​OutsideRuntimeClient.cs	Implements ICallbackDataTarget and performs exact unregister via key/value pair removal.
src/​Orleans.Core/​Runtime/​CallbackData.cs	Adds ICallbackDataTarget and updates callback cleanup paths to call target.Unregister(this).
src/​Orleans.Core/​Messaging/​StripedCallbackDictionary.cs	Adds the striped dictionary implementation, including an “expected value/instance” removal overload and snapshot iteration.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[github-actions]

Code coverage

78.54% line coverage - 99,514 / 126,708 lines

Coverage details

• Source files: 1,663
• Scope: every CI test matrix job, including providers, CodeGen, .NET 8/10, Linux, Windows, and macOS
• Sources: maintained repository code under src/ loaded by the complete CI test scope
• Commit: 9b86a0c3f0
• View the test run and merged coverage artifact
• View the coverage reporting run

[Copilot]

Copilot review overview

Review tier: Lite
Findings: None

Issues resolved since last review (1)

Severity	Finding
	src/​Orleans.Core/​Messaging/​StripedCallbackDictionary.cs — TryRemove(id, expected) uses ReferenceEquals(value, expected), which will always return false… View resolved comment

Suppressed comments (1)

Previously missed (1) — in code that hasn't changed since the last review.

src/Orleans.Core/Messaging/StripedCallbackDictionary.cs:153

• CountWhere executes the predicate while holding the stripe lock. That means an expensive predicate (or one which calls back into this dictionary) can stall other operations and can also deadlock via re-entrancy. Consider snapshotting the stripe values (as ForEach does) and evaluating the predicate outside the lock.

            lock (stripe.Lock)
            {
                foreach (var value in stripe.Dictionary.Values)
                {
                    if (predicate(value))

[Copilot]

Copilot review overview

Review tier: Lite
Findings: 1

New issues introduced by this change (1)

Severity	Finding
	src/​Orleans.Core/​Messaging/​StripedCallbackDictionary.cs — TryRemove(id, expected) currently uses ReferenceEquals(value, expected), which makes the…

[Copilot]

Copilot review overview

Review tier: Lite
Findings: 1

Pre-existing issues (1)

Severity	Finding
	src/​Orleans.Core/​Messaging/​StripedCallbackDictionary.cs — TryRemove(id, expected) currently uses ReferenceEquals(value, expected), which makes the… View comment

Suppressed comments (1)

src/Orleans.Core/Messaging/StripedCallbackDictionary.cs:105

• TryRemove(id, expected) currently uses ReferenceEquals(value, expected). For value-type TValue (or structs), this will always be false (boxing), so exact removal can never succeed. If this dictionary is intended to support non-reference TValue, switch to value equality for value types while keeping reference identity for reference types (so CallbackData exact ownership semantics remain unchanged).

            if (!stripe.Dictionary.TryGetValue(id, out var value) || !ReferenceEquals(value, expected))
            {
                return false;
            }

[Copilot]

Copilot review overview

Review tier: Lite
Findings: None

Issues resolved since last review (1)

Severity	Finding
	src/​Orleans.Core/​Messaging/​StripedCallbackDictionary.cs — TryRemove(id, expected) currently uses ReferenceEquals(value, expected), which makes the… View resolved comment