…c-context logic (#22099)

Removes the ioredis unit test that stubbed the SDK's async context
strategy to drive spans, which replicated SDK internals in the test
layer. Per the review thread in #21885, we rely on the integration tests
instead.

The orchestrion path is covered by the `redis` (ioredis scenario) and
`redis-cache` node integration suites. The `responseHook` and
connect-span cases aren't cleanly testable on the orchestrion path (the
hook is wired internally for caching) and these suites are Docker-gated,
so they're left as a possible follow-up rather than migrated now.

Ref:
#21885 (comment)

Co-authored-by: Opus 4.8 <noreply@anthropic.com>

…nc-context logic (#22101)

Removes the mongoose unit test that stubbed the SDK's async context
strategy to drive spans, which replicated SDK internals in the test
layer. Per the review thread in #21885, we rely on the integration tests
instead.

The path is covered by the `mongoose-tracing-channel` node integration
suite. To avoid losing the useful unit-only assertions, this PR also
adds error-status, empty-filter, and cursor-batchSize coverage to that
suite.

Ref:
#21885 (comment)

---------

Co-authored-by: Opus 4.8 <noreply@anthropic.com>

…-context logic (#22098)

Follow-up to the review discussion in #21885: several `server-utils`
diagnostics-channel unit tests stub the SDK's async context strategy
(`setAsyncContextStrategy` plus a hand-rolled
`installTestAsyncContextStrategy`) to drive spans, replicating SDK
internals in the test layer. Per the thread, we're removing these in
favor of the integration tests that exercise the real code paths.

This PR removes both redis unit tests:
- `test/integrations/tracing-channel/redis.test.ts` (orchestrion path,
origin `auto.db.orchestrion.redis`) — covered by `suites/tracing/redis`
(the suite branches on `isOrchestrionEnabled()`).
- `test/redis/redis-dc-subscriber.test.ts` (native diagnostics-channel
subscriber, origin `auto.db.redis.diagnostic_channel`) — covered by
`suites/tracing/redis-dc`.

Ref:
#21885 (comment)

Co-authored-by: Opus 4.8 <noreply@anthropic.com>

…async-context logic (#22097)

Removes the postgres.js unit test that stubbed the SDK's async context
strategy to drive spans, which replicated SDK internals in the test
layer. Per the review thread in #21885, we rely on the integration tests
instead.

The orchestrion path is already covered end to end by the `postgresjs`
node integration suite. The remaining unit-only assertions checked
internal implementation details (not re-creating a span on a re-entrant
call, skipping an already-wrapped query, no span without an active
parent) rather than the spans a user actually sees, so there was nothing
meaningful to move into an integration test.

Ref:
#21885 (comment)

Co-authored-by: Opus 4.8 <noreply@anthropic.com>

…nc-context logic (#22103)

Removes the three graphql subscriber unit tests and their shared helper,
which stubbed the SDK's async context strategy and a fake document to
drive spans, replicating SDK internals in the test layer. Per the review
thread in #21885, we rely on the integration tests instead.

The path is covered by the `graphql-tracing-channel` node integration
suite. To avoid losing the useful unit-only assertions, this PR also
adds validation-error status and `ignoreTrivialResolveSpans` coverage to
that suite. The subscription-channel case is left as a possible
follow-up.

Ref:
#21885 (comment)

---------

Co-authored-by: Opus 4.8 <noreply@anthropic.com>

…ontext logic (#22100)

Removes the hapi unit test that stubbed the SDK's async context strategy
to drive spans, which replicated SDK internals in the test layer. Per
the review thread in #21885, we rely on the integration tests instead.

The orchestrion path is covered by the `hapi` node integration suite.
`hapi-utils.test.ts` stays (pure metadata formatting, no async-context
stubbing). The remaining unit-only assertions checked internal
implementation details (setup only wrapping once when called twice, no
span without an active span, accepting the ext handler as a tuple or an
object) rather than the spans a user actually sees, so there was nothing
meaningful to move into an integration test.

Ref:
#21885 (comment)

Co-authored-by: Opus 4.8 <noreply@anthropic.com>