interleave data stream

gistfile1.txt

say you have data sources with pagination, where you iterate with page and per_page
http://host/articles?topic=action  (page = [0,1,2,...] per_page = 10 | 20 | 30)
http://host/articles?topic=puzzle  (page = [0,1,2,...] per_page = 10 | 20 | 30)
http://host/articles?topic=shooter (page = [0,1,2,...] per_page = 10 | 20 | 30)
http://host/articles?topic=rpg     (page = [0,1,2,...] per_page = 10 | 20 | 30)

Each article is tagged one and one only for the 'topic'. And then, you
realize that some topics (action, puzzle, shooter, rpg) are actually 'games'
so you define game is a set topics [action, puzzle, shooter, and rpg)

Now how would you interleave specific topic (i.e. action, puzzle...)
for less specific topic (i.e. games)?

If you think this is mistagging issue. Well, it is not. 
This is still a valid problem. What if you have data from twitter, 
facebook, google, etc where you want to aggregate all of them 
for a 'coarse' topic?

How do I have a 'fair share' from each data source while 
honoring the main 'page' and 'per_page' contract? 
From client perspective, I want to look like a 'single' data source

How do I interleave in a such way that you don't show 
too much / too little of one data source?

What if the data source can not guarantee / satisfy your request in whole
For example, if you ask 20 items, it may just return you 18?
Would you fullfil the missing two from other data sources? if so, how?