@dcausse changing how the Wikipedias delta is generated is quite tricky: article-level suggestions are isolated from section-level ones, and the respective DAGs write the delta separately. Hence, we'd need to build a new Spark job + DAG that waits for those 2 deltas and outputs the final one.
Numbers suggest that the amount of page IDs with more than one update is relatively small, compared to the overall amount: the last 5 snapshots have 719,128 page IDs to be updated, of which 22,746 will get more than one update, and 20,828 will get two updates.

Is this a hard requirement for you or a nice to have?

I think this task should be done as part of T372912 which will involve some refactoring of the way the tags are shipped.
I suspect that the delta you generate could easily be grouped by page_id after they're computed.

The impact of having multiple updates per article:

• greater number of updates (which we want to minimize for T372912)
• conflicts, perhaps rare but if two updates for the same page_id are sent close to each other one might fail on a conflict when entering elastic.