# Description
https://github.com/cowprotocol/services/pull/2906 moved refund indexer
maintenance into the solvable orders cache update loop. When there are
no ethflow refunds in the DB or the last refund is very old, the refund
indexer tries to fetch all the blocks, which might execute for more than
10 minutes, resulting in autopilot restarts and the whole protocol being
stuck.

# Changes

Change the behavior of calculating the starting index for refund
indexer. If a recent eth-flow refund exists within the last day, it
prioritizes this. Otherwise, it falls back to the most recent block from
broadcasted orders. Then select the highest value among the value
calculated in the previous step, settlement contract block and
configured `ethflow_indexing_start`.

## How to test
Staging.
<img width="598" alt="image"
src="https://github.com/user-attachments/assets/fed96a1b-2523-4ca5-9fb4-767b8c87e536">

(cherry picked from commit 68815a36b83ae32de46df09f3c8a203e5394b567)