…#39309)

## **Description**

`ShieldSubscriptionProvider` was creating a new context value object on
every render and recreating `evaluateCohortEligibility` whenever any of
its 10 dependencies changed, causing unnecessary re-renders of the Home
component subtree.

Two fixes applied:

1. **Stabilized callback with ref pattern** -- Moved the
`evaluateCohortEligibility` implementation into a ref updated via
`useEffect`. The public callback delegates to the ref, keeping a stable
reference (`useCallback` with empty deps) while still accessing current
closure values.

2. **Memoized context value** -- Wrapped the provider value object in
`useMemo` keyed on the now-stable callback, preventing a new object
reference on every render.

This is fix 6 of 6 in the [P0 Global Cascade Re-render Fixes
epic](https://github.com/MetaMask/MetaMask-planning/issues/6669)
([RCA](https://docs.google.com/document/d/1KrzqL1IEQb-ctOKpOxhjBVr_-5jDXTgwC8a4r1ICBg0/edit?tab=t.5g1bh4tznel0),
[Profiling
Report](https://docs.google.com/document/d/1zlTg8rWYUFKZAsjQF3I9o7KaMA6uq7qcYjdfWKIZIJU/edit?tab=t.0#heading=h.dxmx8vunwu2s)).
Together, the 6 memoization fixes break the cascade chain originating in
Routes/Home that propagated through the entire component tree on every
navigation and state update.

[![Open in GitHub
Codespaces](https://github.com/codespaces/badge.svg)](https://codespaces.new/MetaMask/metamask-extension/pull/39309?quickstart=1)

## **Profiling Results** (all 6 fixes combined -- [Full
Report](https://docs.google.com/document/d/1zlTg8rWYUFKZAsjQF3I9o7KaMA6uq7qcYjdfWKIZIJU/edit?tab=t.0#heading=h.dxmx8vunwu2s))

We ran 7 rounds of controlled A/B benchmarks over 41 hours. Results
confirm the cascade was the dominant source of interactive UI latency.
WDYR audit: 0 cascade re-renders on treatment (RCA measured 125-200 per
action on baseline).

> Measured on test builds (`yarn build:test`) with React StrictMode,
Power User persona (30 accounts). All p-values from Welch's t-test,
confirmed by Mann-Whitney U. Effect sizes: Cohen's d = 0.93-5.11 (large
to very large). n=5-20 per metric.

### Per-interaction improvements

| Flow | Baseline | Treatment | Improvement | p |
|------|----------|-----------|-------------|---|
| Account switching (5 accts) | 3.1s | 1.7s | **-1.4s, 47% faster** |
0.044 |
| Asset detail nav (ETH) | 83ms | 37ms | **-46ms, 55% faster** | 0.044 |
| Asset detail nav (SOL) | 75ms | 31ms | **-44ms, 59% faster** | <0.001
|
| Send flow (e2e) | 928ms | 804ms | -124ms, 13% faster | 0.044 |
| Send: token select to form | 25ms | 13ms | **-12ms, 49% faster** |
0.040 |

### Cascade compounding under sustained interaction

Route cycling benchmark (3 round-trips: Send -> Assets -> Swap):

| Segment | Baseline | Treatment | Improvement | p |
|---------|----------|-----------|-------------|---|
| **Total (6 transitions)** | **19.9s** | **2.7s** | **-17.2s, 7x
faster** | 0.004 |
| sendToHome | 6.4s | 723ms | -5.7s, 89% faster | 0.018 |
| assetDetailsToHome | 6.3s | 659ms | -5.7s, 90% faster | 0.004 |
| swapToHome | 5.0s | 787ms | -4.2s, 84% faster | 0.001 |
| homeToSwap | 877ms | 227ms | -650ms, 74% faster | 0.002 |
| homeToAssetDetails | 628ms | 236ms | -392ms, 62% faster | 0.004 |
| homeToSend | 673ms | 82ms | -591ms, 88% faster | 0.046 |

The 7x is a **compounding effect** across successive transitions, not a
per-transition effect. Isolated transitions show only 10-15%
improvement. Each navigation accumulates subscription/listener state on
baseline, escalating re-render overhead -- consistent with the RCA's
cascade chain model (Routes -> Home -> all children).

This compounding behavior is itself a key finding: it means any
remaining or future cascade leak will degrade **non-linearly** with
session length, even if it looks trivial in a single benchmark.

**Production estimate:** StrictMode roughly doubles render work. The 7x
measured in dev builds is expected to land at ~2-7x in production
depending on wallet size, tree depth, and GC timing.

### Metrics to watch after merge

- **CI (`metamask-performance`):** `assetClickToPriceChart` should show
a step-change (83->37ms dev). `cascadeFixProfiling` preset can be added
for ongoing regression tracking.
- **Production Sentry (RUM):** Best signal at **p75/p95 navigation
latency** and **INP p95**, segmented by users with 10+ accounts. Expect
improvement on route transitions, account switching, token list
rendering. Signal will be diffuse due to network variance and device
heterogeneity -- don't expect a clean step-function.
- **TBT:** Cascade generates clusters of consecutive long tasks (>50ms)
on every state change. Fix eliminates these. Should show up in Sentry
performance dashboard once long-task tracking is enabled.

## **Changelog**

CHANGELOG entry: Fix memoization issue in top-level context provider
that was causing cascading re-renders.

## **Related issues**

Fixes: MetaMask/MetaMask-planning#6638

Part of epic: [MetaMask-planning#6669 -- P0 Global Cascade Re-render
Fixes](https://github.com/MetaMask/MetaMask-planning/issues/6669)

Sibling PRs:
- [#39310](#39310) --
Memoize MetaMetrics context
([#6640](https://github.com/MetaMask/MetaMask-planning/issues/6640))
- [#39311](#39311) --
Stabilize `withRouterHooks` params/location
([#6671](https://github.com/MetaMask/MetaMask-planning/issues/6671))
- [#39312](#39312) --
Fix `getPendingApprovals` selector
([#6673](https://github.com/MetaMask/MetaMask-planning/issues/6673))
- [#39313](#39313) --
Fix `pendingConfirmationsSortedSelector`
([#6674](https://github.com/MetaMask/MetaMask-planning/issues/6674))
- [#39314](#39314) --
Remove dead What's New modal code
([#6670](https://github.com/MetaMask/MetaMask-planning/issues/6670))

## **Manual testing steps**

## **Screenshots/Recordings**

## **Pre-merge author checklist**

- [x] I've followed [MetaMask Contributor
Docs](https://github.com/MetaMask/contributor-docs) and [MetaMask
Extension Coding
Standards](https://github.com/MetaMask/metamask-extension/blob/main/.github/guidelines/CODING_GUIDELINES.md).
- [x] I've completed the PR template to the best of my ability
- [x] I've included tests if applicable
- [x] I've documented my code using [JSDoc](https://jsdoc.app/) format
if applicable
- [x] I've applied the right labels on the PR (see [labeling
guidelines](https://github.com/MetaMask/metamask-extension/blob/main/.github/guidelines/LABELING_GUIDELINES.md)).
Not required for external contributors.

## **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.


<!-- CURSOR_SUMMARY -->
---

> [!NOTE]
> **Medium Risk**
> Changes how `evaluateCohortEligibility` is wired (ref-backed stable
callback) and how the context value is memoized, which could alter
when/with-what state eligibility checks run if the ref is not kept in
sync. Added tests reduce risk but this touches subscription/cohort
gating that affects modal display behavior.
> 
> **Overview**
> Reduces cascading re-renders from `ShieldSubscriptionProvider` by
making the exported `evaluateCohortEligibility` function
reference-stable and memoizing the context `value` object.
> 
> Refactors `evaluateCohortEligibility` to delegate through a
`useRef`-stored implementation (updated during render so it’s available
in commit-phase callbacks) while keeping the public callback
`useCallback([])`. Adds a new Jest test suite validating child
rendering, stable callback/context identity across re-renders, and that
the stable callback still observes updated selector values.
> 
> <sup>Written by [Cursor
Bugbot](https://cursor.com/dashboard?tab=bugbot) for commit
50a2fed. This will update automatically
on new commits. Configure
[here](https://cursor.com/dashboard?tab=bugbot).</sup>
<!-- /CURSOR_SUMMARY -->

---------

Co-authored-by: Jongsun Suh <jongsun.suh@icloud.com>