statement-store: close light node test coverage gaps

[AndreiEres]

Summary

Tracking issue for the remaining scenarios in statement-store light node tests. See details in the testing strategy.

Tasks

P0

• Wire protocol: unit tests cover edge cases around truncated payloads and compact encoding boundaries, polkadot-sdk cross-compatibility, can be extended with fuzzing statement-store: Add wire protocol V1 encode/decode edge case tests #3187
• Submission: unit tests cover result mapping, Zombienet tests verify statements sent by the light node reach connected full nodes
  • Unit statement-store: Add unit tests for statement submission and reception #3188
  • Zombienet statement-store: Add zombienet test #3191
• Reception: unit tests cover bloom filter construction, false-positive discard, deduplication, subscription notifications, and unsubscribe; Zombienet tests verify only subscribed statements are delivered without duplicates
  • Unit statement-store: Add unit tests for statement submission and reception #3188
  • Zombienet statement-store: Add zombienet test #3191
• Peer connection: Zombienet tests verify connectivity is maintained or restored when full nodes drop statement-store: Add zombienet test #3191

P1

• Long-session stability: Versi tests verify a light node remains stable while connected to real full nodes for hours [DNM] Add statement-latency-bench example #3239
• Browser sanity check: Zombienet tests verify the basic flow works in a headless or real browser environment statement-store: Add browser sanity check zombienet test #3244

[AndreiEres]

Long-session stability tests failed.

The goal was to run 500 smoldot-based clients at load ≈540 000 st/hr and measure latency. But that isn't possible with smoldot today: the WASM client starts panicking once more than ~100 clients are spawned (across separate pods). At this scale it is not stable enough for long sessions.

The testing work that this issue was tracking is done, so I'm closing it. The follow-up — switching from “confirm long-session stability” to “find the circumstances under which smoldot loses stability” — will be tracked in a separate issue.

cc @P1sar @alexggh