test: reduce flakiness in Copilot Chat sanity tests

[ulugbekna]

The Copilot Chat sanity tests in extensions/copilot hit the live model endpoint, so they can fail for transient upstream reasons (empty response, intermittent rate limiting, model errors) that do not represent an actual regression.

Most recent example: build 444205 failed at macOS (ARM64) attempt 1, where Slash Commands work properly hit:

AssertionError: ok(progressReport.currentProgress)

The /explain request returned, but currentProgress (which only accumulates MarkdownPart/AnchorPart text) was empty. With retries: 1 already configured, both attempts failed and we had zero diagnostics to tell a flake from a real bug.

Changes

• extensions/copilot/.vscode-test.mjs — bump mocha retries for sanity tests from 1 to 2 (3 attempts total). The suite-level timeout is 1 min per test, and the pipeline job has a 30 min timeout, so this is well within budget.
• extensions/copilot/src/extension/test/vscode-node/sanity.sanity-test.ts — add a shared describeOutcome(stream, result?) helper used in every assert.ok(currentProgress, …) call site. On failure the assertion message now includes:
  • the stream's item count and the constructor names of each part (so we can see if only non-markdown parts came back),
  • the actual currentProgress value,
  • result.metadata.responseId for correlating with server-side logs,
  • result.errorDetails if the chat handler reported an upstream error (rate limit, content filter, server error — these are returned via the result, not thrown).

Also reuses the helper in the existing agent-mode timeout diagnostics instead of duplicating the dump logic.

Verification

• tsgo --noEmit --project tsconfig.json in extensions/copilot passes.
• Pre-commit hygiene hook passes.

cc @lramos15 @roblourens

Related: #312131, #313303, #317255

[Copilot]

Pull request overview

Reduces flakiness in the Copilot Chat sanity test suite (which hits live model endpoints) by increasing retry budget and improving failure diagnostics so CI logs can distinguish transient upstream failures from real regressions.

Changes:

• Increase Mocha retries for sanity runs from 1 to 2 (3 total attempts).
• Add a shared describeOutcome(stream, result?) helper and use it in assertion messages and agent-mode timeout errors to include stream-part breakdown, currentProgress, responseId, and errorDetails.

Show a summary per file

File	Description
extensions/copilot/src/extension/test/vscode-node/sanity.sanity-test.ts	Adds/uses describeOutcome to enrich assertion and timeout diagnostics for flaky upstream failures.
extensions/copilot/.vscode-test.mjs	Increases Mocha retries for sanity tests to allow additional attempts against transient endpoint issues.

Copilot's findings

• Files reviewed: 2/2 changed files
• Comments generated: 5