From the code in OnGetPagesByURLDone from offline_page_utils.cc [1], when selecting among many matches of offline pages for a given URL, higher preference is given to the most recently created one. This logic might be favoring the loading of offline pages generated by last_n because of its "aggressiveness" in generating them.

The final effect I am concerned about is that usage metrics would be reporting accesses to the last_n namespace over other namespaces. This would not necessarily be a bug but it all depends on the specific use case. This is one bad example I can think of:

* A suggested article offline page is made initially available by the prefetching system.
* User opens the NTP and sees that article's entry and *because it is badged* decides to open it.
* As the user is currently online, the actual online version of the page is loaded.
* User switches apps and so last_n generates a new offline page.
* Later on the user is offline and relaunches Chrome.
* The more recent last_n offline page takes preference and is loaded instead of the prefetched one.

In this case, even though it was the prefetching feature that motivated the opening of that offline page the "point" would go to last_n. There's also a potential page quality issue: even thought the last_n is more recent, server generated pages are more likely to have higher fidelity to the original one.

Limiting factors: last_n pages can only be loaded in the very tab they were created and they are limited the latest loaded page in that tab. These could make this a non-issue but I'm unsure about that.


[1] https://cs.chromium.org/chromium/src/chrome/browser/offline_pages/offline_page_utils.cc?q=OfflinePageUtils::SelectPageForURL%7COnGetPagesByURLDone&dr=C