)]}'
{"test/probe/test_account_failures.py":[{"author":{"_account_id":15343,"name":"Tim Burke","email":"tburke@nvidia.com","username":"tburke"},"change_message_id":"4be0a300d18aec74de402700bef740708957efd5","unresolved":true,"context_lines":[{"line_number":104,"context_line":"        client.delete_container(self.url, self.token, container1)"},{"line_number":105,"context_line":""},{"line_number":106,"context_line":"        # Container DELETEs distribute account updates across container"},{"line_number":107,"context_line":"        # backends, so up to one account primary may miss the update."},{"line_number":108,"context_line":"        # Converge the delete before creating another outstanding update."},{"line_number":109,"context_line":"        Manager([\u0027account-replicator\u0027]).once()"},{"line_number":110,"context_line":""}],"source_content_type":"text/x-python","patch_set":1,"id":"01292e13_c692aa38","line":107,"range":{"start_line":107,"start_character":49,"end_line":107,"end_character":68},"updated":"2026-06-03 03:54:18.000000000","message":"...because taking down the account disk (may have) also taken down a primary container replica, is that right?\n\nWould it be better to repeatedly pick random container names until we find one whose primaries have no overlap with the down disk?","commit_id":"574d5513dc7e63909e36d93c8f136e9f2419cd00"}]}