When streaming mode is enabled or the current span is a StreamedSpan, `get_start_span_function()` returns `sentry_sdk.traces.start_span` which accepts `(name, attributes, parent_span)`. However, all callers (e.g., anthropic.py:408, google_genai, langchain, litellm, mcp, openai_agents) invoke it with `(op=..., name=..., origin=...)`. This will cause a `TypeError` at runtime because `op` and `origin` are not accepted parameters, and `name` is a positional-only argument in the new API.

When streaming mode is enabled or the current span is a StreamedSpan, `get_start_span_function()` returns `sentry_sdk.traces.start_span` which accepts `(name, attributes, parent_span)`. However, all callers (e.g., anthropic.py:408, google_genai, langchain, litellm, mcp, openai_agents) invoke it with `(op=..., name=..., origin=...)`. This will cause a `TypeError` at runtime because `op` and `origin` are not accepted parameters, and `name` is a positional-only argument in the new API.

In `_sentry_execute_command`, the `db_span` and `cache_span` are opened via `__enter__()` (lines 120, 135) but their `__exit__()` calls (lines 142, 146) are not protected by try/finally. If `await old_execute_command()` raises an exception, the spans will never be closed, causing resource leaks and incorrect span hierarchy. The sync version in `_sync_common.py` correctly uses `try...finally` to ensure spans are always closed.

In `_sentry_execute_command`, the `db_span` and `cache_span` are opened via `__enter__()` (lines 120, 135) but their `__exit__()` calls (lines 142, 146) are not protected by try/finally. If `await old_execute_command()` raises an exception, the spans will never be closed, causing resource leaks and incorrect span hierarchy. The sync version in `_sync_common.py` correctly uses `try...finally` to ensure spans are always closed.

In the span streaming branch, `self.graphql_span` is created via `sentry_sdk.traces.start_span()` at line 192 but is never started with `span.start()` or used as a context manager. When `self.graphql_span.end()` is called at line 237, it calls `__exit__()` which tries to access `self._context_manager_state` that was only set in `__enter__()`. Since the span was never entered, this raises an AttributeError that gets silently caught by `capture_internal_exceptions()`. The span will never be properly sent to Sentry.

In the span streaming branch, `self.graphql_span` is created via `sentry_sdk.traces.start_span()` at line 192 but is never started with `span.start()` or used as a context manager. When `self.graphql_span.end()` is called at line 237, it calls `__exit__()` which tries to access `self._context_manager_state` that was only set in `__enter__()`. Since the span was never entered, this raises an AttributeError that gets silently caught by `capture_internal_exceptions()`. The span will never be properly sent to Sentry.

In the span streaming branch, `self.graphql_span` is created via `sentry_sdk.traces.start_span()` at line 192 but is never started with `span.start()` or used as a context manager. When `self.graphql_span.end()` is called at line 237, it calls `__exit__()` which tries to access `self._context_manager_state` that was only set in `__enter__()`. Since the span was never entered, this raises an AttributeError that gets silently caught by `capture_internal_exceptions()`. The span will never be properly sent to Sentry.

The changes add new code branches for handling `StreamedSpan` in streaming mode (lines 215-221, 231-237, 245-255, 265-266, 276-285), but the existing tests only exercise the legacy `Span` code path via `start_transaction`. The new streaming mode paths (`_experiments={"trace_lifecycle": "stream"}`) have no test coverage in the rust_tracing integration tests.

`NoOpStreamedSpan.__init__` sets `self.segment = None`, but it does not override inherited methods like `dynamic_sampling_context()`, `to_baggage()`, `iter_headers()`, and `get_baggage()` which all access `self.segment.get_baggage()`. Calling any of these methods on a `NoOpStreamedSpan` instance will raise `AttributeError: 'NoneType' object has no attribute 'get_baggage'`.

`NoOpStreamedSpan.__init__` sets `self.segment = None`, but it does not override inherited methods like `dynamic_sampling_context()`, `to_baggage()`, `iter_headers()`, and `get_baggage()` which all access `self.segment.get_baggage()`. Calling any of these methods on a `NoOpStreamedSpan` instance will raise `AttributeError: 'NoneType' object has no attribute 'get_baggage'`.