fix: soften await parallelization messages to acknowledge uncertainty

[skoshx]

Summary

Softens the messaging in server-sequential-independent-await and async-await-in-loop to acknowledge that parallelization only helps when work is truly independent—no shared queues, transactions, ordering requirements, or synchronous facades behind async interfaces.

Root Cause

The rules made categorical claims about performance benefits:

• "your users wait twice as long for nothing"
• "This makes the loop slow"

These claims don't hold when:

1. Work is serialized through a shared queue (SQLite connection, Redis client, transaction manager)
2. Async functions are facades over synchronous work (Set.has behind async exists())
3. Operations maintain ordering guarantees by design

The detector cannot prove runtime independence from syntax alone.

Changes

Messaging updates:

• server-sequential-independent-await: "These awaits appear independent, but parallelization only improves performance when work doesn't share queues, transactions, or resources. Verify independence, then consider Promise.all([...])."
• async-await-in-loop: "This loop runs awaits sequentially. Parallelization may improve performance if work is truly independent (no shared queues, transactions, or ordering requirements). Verify before applying..."

Detection unchanged:

• No new exemptions added
• Rules still fire on the same patterns
• Added regression tests for serialized queues and sync facades (async await diagnostics claim parallelism benefits for serialized or synchronous work #1840)

Scope Decision

Per the issue request, this fix addresses messaging appropriateness, not detection scope. The rules cannot reliably detect shared queues or synchronous facades through static analysis, so the solution is conditional language that asks developers to verify independence rather than claiming definite latency improvements.

Blanket suppression based on method names was explicitly rejected as it would create false negatives.

Testing

• All existing tests pass
• New regression tests confirm rules still fire on reported patterns
• Reproduction case shows updated messages
• Lint and typecheck clean

Parity

Parity was not run because:

1. Changes are purely to message strings
2. No detection logic was modified
3. Every diagnostic from these rules will show a "diff" (the message text)
4. That message change is the entire point of this fix

Closes #1840

  

[pkg-pr-new]

Open in StackBlitz

npm i https://pkg.pr.new/eslint-plugin-react-doctor@1841

npm i https://pkg.pr.new/oxlint-plugin-react-doctor@1841

npm i https://pkg.pr.new/react-doctor@1841

commit: 0283f56

[github-actions]

Interactive terminal E2E

Terminal Control verified the built CLI at 0283f56 in a real PTY:

• selected a project interactively and observed Scanning... before the three-second Git delay completed
• waited for the clean result and exercised the compact report
• opened copy context and the GitHub Actions confirmation, then cancelled safely

Download the edited MP4 and PNG evidence