Yochana Flow responses can contain items nested inside objects and arrays at unpredictable depths. Write a function that recursively extracts valid items, removes duplicate IDs by keeping the most recently updated version, normalizes their fields, and returns them in a deterministic order.
An object is a valid item when it contains non-empty string fields id and title. Optional fields are subtitle, imageUrl, and updatedAt. Missing optional fields must become empty strings in the output. The updatedAt value, when present, is an ISO 8601 UTC string, so lexicographic comparison is valid.
Input response is a JSON-compatible Python value consisting only of dictionaries, lists, strings, numbers, booleans, and None. Return a list of dictionaries with exactly these keys: id, title, subtitle, image_url, and updated_at.
If multiple items have the same id, keep the item with the lexicographically greatest updatedAt. Treat a missing updatedAt as an empty string. Sort the final result by updated_at descending, then by id ascending. If duplicate items have equal timestamps, keep the first one encountered during traversal.
def parse_y_flow_response(response):