Description

The Hyperliquid SDK update from 0.30.2 to 0.32.2 (PR #28672, shipped in 7.74.0) brought in @nktkas/rews v2, which now properly surfaces reconnect failures that v1 silently swallowed. Combined with the new spotState WS path from the 7.73.2 cherry-picks (b5947ba9f9, 985830d501), this produced ~15k Sentry events/week on 7.74.1 from errors that the SDK already catches and recovers from automatically.

The events have no user-visible impact: they're tagged mechanism: generic, handled: yes and the SDK reconnects on its own. They drown out genuine signal like Failed to fetch market data (which we want to keep capturing).

This PR generalizes the existing #isTransientAssetCtxsError helper in HyperLiquidSubscriptionService to #isTransientSdkError, broadens its matching to cover all four transient SDK error classes, and gates #logErrorUnlessClearing so transient errors are downgraded to debugLogger.log (kept for local diagnosis) instead of forwarded to Sentry.

Single chokepoint — all four affected catch sites (refreshSpotState, createUserDataSubscription webData3 / HIP-3 / webData2) already routed through #logErrorUnlessClearing, so no per-site changes are required.

Suppressed error classes (only when caught and recovered by the SDK):

Still captured:

• Non-transient errors (TypeError, RangeError, etc.)
• Failed to fetch market data — wsState=disconnected (5GMV) — real downstream user-visible signal
• CLIENT_NOT_INITIALIZED (5R41) — real init failures
• ValiError (TAT-3093 path, fixed separately by fix(perps): non-EVM address passed to HyperLiquid validator via usePerpsPositionForAsset #29420)
• AbortError on candle fetch — already filtered by fix(perps): historical-candle cancellation handling  #28953 + fix: Perps chart viewport breaking when revisiting a candle interval #29344, awaiting release

The retry decision in createAssetCtxsSubscription now uses the same broader matcher. This means the assetCtxs path may retry on slightly more error types (TimeoutError, ReconnectingWebSocketError) before giving up — equal-or-more retries, never less.

Changelog

CHANGELOG entry: null

Related issues

Related Sentry issues: METAMASK-MOBILE-5SQY, METAMASK-MOBILE-5SEB, METAMASK-MOBILE-5RV9, METAMASK-MOBILE-5SG9, METAMASK-MOBILE-5SGD, METAMASK-MOBILE-5SGK, METAMASK-MOBILE-4FE2

Manual testing steps

Feature: Transient Hyperliquid SDK errors do not pollute Sentry

  Scenario: User briefly opens Perps on a flaky network
    Given the wallet is unlocked
    And Perps is enabled
    When the user enters Perps and the WebSocket connection fails or times out
    Then the SDK reconnects automatically
    And the transient WebSocketRequestError, ReconnectingWebSocketError, or TimeoutError is logged to DevLogger only
    And no Sentry event is captured for these classes
    And the user-visible "Failed to fetch market data" error (when sustained) still captures to Sentry

  Scenario: User triggers a non-transient error path
    Given the wallet is unlocked
    And Perps is enabled
    When a non-transient error (e.g. CLIENT_NOT_INITIALIZED, TypeError) occurs in HyperLiquidSubscriptionService
    Then the error is captured to Sentry as before
    And the existing context tags (feature: perps, service: HyperLiquidSubscriptionService) are preserved

Verification:

• yarn jest app/controllers/perps/services/HyperLiquidSubscriptionService.test.ts --no-coverage — 118 passed, 4 skipped, 0 failed
• npx eslint app/controllers/perps/services/HyperLiquidSubscriptionService.ts — clean
• NODE_OPTIONS='--max-old-space-size=8192' npx tsc --noEmit --incremental — exit 0
• For weak-network manual repro, use Network Link Conditioner (iOS) or adb shell svc data disable && sleep 3 && adb shell svc data enable (Android) while in Perps; verify Metro log shows [Perps transient SDK error] entries and Sentry receives no new events for the listed classes.

Screenshots/Recordings

N/A — internal observability change with no UI surface. Verification is via Sentry dashboard delta after release.

Before

After 7.74.1 release, feature:perps Sentry volume on Android: ~77k events / week. ~15k/wk attributable to the suppressed transient classes (in addition to ~54k AbortError already filtered by #28953 + #29344, pending release).

After

Expected feature:perps Sentry volume after this PR + the unreleased AbortError fixes ship: ~4–8k events / week (close to pre-7.74.x baseline), composed of genuine error classes only.

Pre-merge author checklist

• I've followed MetaMask Contributor Docs and MetaMask Mobile Coding Standards.
• I've completed the PR template to the best of my ability
• I've included tests if applicable — existing 118 tests cover the regression case (non-transient errors still capture); new transient classes share the same code path that the existing tests exercise.
• I've documented my code using JSDoc format if applicable
• I've applied the right labels on the PR (see labeling guidelines). Not required for external contributors.

Performance checks (if applicable)

• I've tested on Android — N/A (no runtime behavior change for users; only changes which errors reach Sentry)
• I've tested with a power user scenario — N/A
• I've instrumented key operations with Sentry traces for production performance metrics — N/A; this PR removes noise from Sentry rather than adding traces.

Pre-merge reviewer checklist

• I've manually tested the PR (e.g. pull and build branch, run the app, test code being changed).
• I confirm that this PR addresses all acceptance criteria described in the ticket it closes and includes the necessary testing evidence such as recordings and or screenshots.