fix: fallback token holders for hot tokens

[o-az]

Summary

• avoid expensive full holder aggregation for high-activity tokens
• add a recent-participant fallback that reads current balanceOf values and sorts positive balances
• update holder balance aggregation tests for split incoming/outgoing queries

Validation

• pnpm check:types
• pnpm vitest run test/tempo-queries.test.ts -t "fetchTokenHolderBalances"
• pnpm check:env
• pnpm check:biome

[github-actions]

Bundle Size Report

Metric	Size	Δ Change
Total	4.3 MB	+69.0 KB (+1.6%)
Gzip	1.2 MB	+21.1 KB (+1.7%)
Brotli	1.1 MB	+18.5 KB (+1.7%)

Chunk changes (>1KB)

Chunk	Change
assets/AbiItem.js (removed)	-27.1 KB
assets/tempo-address.js	+1.1 KB
assets/_address.js	+1.2 KB
assets/TxTransactionRow.js	+1.5 KB
assets/token.js	+1.6 KB
assets/tempo-queries.js	+2.4 KB
assets/schemas.js	+7.5 KB
assets/parseAbiItem.js (new)	+25.0 KB
assets/wagmi.config.js	+27.1 KB
assets/queries.js	+27.2 KB

Compared against main branch (baseline from 4/27/2026, 4:02:54 PM)

[github-actions]

Cloudflare Deployments

App	Environment	Status	Preview
explorer	devnet	[OK] Deployed	-
explorer	mainnet	[OK] Deployed	-
explorer	testnet	[OK] Deployed	-
fee-payer	devnet	[>>] Skipped	No changes
fee-payer	moderato	[>>] Skipped	No changes
fee-payer	privy	[>>] Skipped	No changes
og	-	[>>] Skipped	No changes
perf	-	[>>] Skipped	No changes
tokenlist	-	[>>] Skipped	No changes

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: ceede74483

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[o-az]

Follow-up pushed in 0024bee.

Addresses Codex review comments:

• Recent participant lookup failures now retry full holder aggregation before falling back to an empty response.
• balanceOf failures now throw instead of silently zeroing holders/caching bad data.

Also kept RPC changes local to Explorer: server-side wagmi now uses the proxy RPC URL with key plus the normal RPC URL with key via the existing load balancer/rate limiter; no shared rpc-utils package changes.

[o-az]

@codex review

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 0024bee189

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Do not skip full holder aggregation on activity heuristic

This branch turns the “recent participant” path into the primary path whenever activity is high, so allHolders is built from a small candidate set instead of the full transfer-derived holder set. Because candidates are limited to recent addresses, tokens with many older/inactive holders will return incomplete holder lists and inaccurate totals/pagination metadata even when fetchTokenHolderBalances would succeed, which is a correctness regression rather than a true fallback.

Useful? React with 👍 / 👎.

[o-az]

Addressed discussion_r3152152874 in 47987c3. The holders endpoint now tries full transfer-derived aggregation first on cache miss, and only uses the recent-participant balanceOf path if full aggregation fails. That keeps recent participants as a true fallback rather than using activity as a correctness-changing heuristic.

Verified with:

• pnpm check:types
• pnpm check:biome
• pnpm vitest run test/tempo-queries.test.ts -t fetchTokenHolderBalances
• pnpm check:env

Also rechecked the PathUSD holders URL locally; it still renders holder rows via fallback when full aggregation fails.

[o-az]

Also fixed the noisy unverified contract source response in f9d8db4. /api/code now maps upstream 404/unverified source to 204 No Content with a short cache, and the client treats both 204 and legacy 404 as no source (null).

Verified with:

• pnpm check:types
• pnpm check:biome
• curl against the PathUSD URL now returns HTTP 204 instead of 404.

[o-az]

Follow-up for PathUSD holder totals pushed in b55a0aa. The fallback path is not an exact holder-count source; it samples recent transfer participants and reads balances because full TIDX aggregation is failing. So the UI should not render 24, 26, or > 26 as a total.

Now the holders response carries totalKnown; when full aggregation fails and we use the recent-participant fallback, the page hides the holder total/count and uses indefinite pagination (1 of …) instead.

Also checked kitty pane 9: it is still showing TIDX 422s for the full holder aggregation (group by to/from) plus the existing workerd cross-request promise warnings.