fix: WebSocket connection reset on test teardown in blocked-peers test

[sanity]

Problem

test_ping_blocked_peers_solution completes successfully (all contract operations succeed, state propagates correctly), but fails with a WebSocket protocol error during teardown:

Connection reset without closing handshake

Context

This issue was discovered while fixing two related bugs:

• PR fix: propagate connection errors for blocked peer fallback (#2021) #2105: Connection error propagation for blocked peers
• Stacked work: PUT/subscribe parent-child completion (branch fix/put-subscribe-completion)

After fixing the PUT completion and counter overflow bugs, the test's functional requirements all pass:

• ✅ Contract PUT succeeds
• ✅ Nodes GET the contract despite being mutually blocked
• ✅ Subscribe operations complete
• ✅ State updates propagate through gateway
• ✅ All nodes converge to consistent state

The WebSocket reset occurs after operations complete, suggesting a teardown/cleanup issue rather than a functional bug.

Investigation Findings

From previous debugging (see /tmp/blocked-peers-final.txt):

• Counter overflow spam is gone after DashMap fixes
• Test runs to completion functionally
• WebSocket reset happens late in execution
• Likely related to connection lifecycle/cleanup during test shutdown

Possible Causes

1. Test teardown race - WebSocket clients closing before nodes fully shut down
2. Connection cleanup timing - Transport layer closing connections while WS still active
3. Pre-existing issue - May have existed but was masked by other failures

Next Steps

1. Add detailed logging around WebSocket lifecycle (connect, close, shutdown)
2. Check if issue reproduces in other WebSocket-based tests
3. Investigate connection cleanup order during node shutdown
4. Consider if this is test-specific or affects production scenarios

Test Status

Test is currently marked #[ignore] with TODO-MUST-FIX to unblock PR merges for the functional fixes. The test should be re-enabled once the WebSocket issue is resolved.

Related

• PR fix: propagate connection errors for blocked peer fallback (#2021) #2105: Connection error propagation
• Issue Observing high send packet rate #2092: Reserved connection counter underflow (fixed as part of stacked work)

[AI-assisted - Claude]

[github-actions]

🤖 Auto-labeled

Applied labels:

• T-bug (95% confidence)
• P-medium (85% confidence)
• E-medium (80% confidence)
• A-networking (92% confidence)
• A-developer-xp (75% confidence)

Reasoning: The report describes a WebSocket protocol error (connection reset without closing handshake) occurring during test teardown, indicating a bug in connection lifecycle/cleanup rather than a new feature or documentation issue. It impacts tests/CI and developer workflow (tests are currently ignored to allow merges), so priority is medium. The fix likely requires moderate investigation into transport/WS shutdown ordering and test teardown sequencing (E-medium). Relevant areas are networking (WebSocket/transport cleanup) and developer experience (test teardown and CI stability).

Previous labels: none

If these labels are incorrect, please update them. This helps improve the auto-labeling system.