_get_current_turn_contents picks wrong anchor when single_turn agents run in parallel

Finding: In an external report against google-adk (reporter states v2.11.0), two single_turn sub-agents running in parallel on the same session can end up with one agent's output in the other's context, apparently through _get_current_turn_contents (https://github.com/google/adk-python/issues/7488). Limitation: this is an external report; KFA did not reproduce it, and the PoC zip was not inspected. Reported mechanism (HYPOTHESIS from the reporter, not maintainer-confirmed): the reverse scan in _get_current_turn_contents starts at len(events)-1, the tail of the shared session event list. If sub_agent_a finishes first and appends its event, sub_agent_b's scan begins at A's event. It skips that event because the branch does not match, then anchors on the parent event directly above it. _get_contents is then called on events[anchor:], which includes A's output. The reporter says ADK wraps the content in a safety block, but the content still reaches B's context (issue #7488). Reported reproduction: the issue says a script, parallel_agent_context_contamination.py, needs no API key, model or LiteLLM. Its reported output shows a scan log in which sub_agent_b's context contains sub_agent_a's event. The issue also says an earlier report (#5961) was closed as stale without a repro, and that PR #6047 (v2.11.0) fixed a different include_contents code path and left this scan untouched. These are the reporter's statements; no other source was supplied to verify them. Status: the issue is open with no maintainer response. A non-maintainer commenter offers to implement the reporter's proposed fix, which snapshots len(session.events) when each sub-agent's InvocationContext is created and uses it as the scan start index. No PR, fix commit or released fix is linked in the supplied sources, so the fix status is proposed only. The issue states no workaround. Actionable implication, as a conditional: anyone running parallel single_turn sub-agents on a shared session who sees unexpected sibling content in context should compare their setup with the issue's described scenario. The sources do not establish a verified mitigation.