get_entries(sort='recent') "denial of service"

[lemon24]

The new heuristic from #279 is too lax, if a feed adds all lots of new entries, it can spam the top of "recent"; this should not be possible.

Background: https://www.michaelnygard.com/atom.xml in my personal instance just added 60+ (years) old entries; for now I marked everything as read, so the added time is preserved for debugging.

Possible solutions:

• Limit how many entries get recent_sort set to "now"; we shouldn't do it based on the entry date, because that would be functionally reverting #279.
  • Is some kind of backfilling needed for this? (I don't need it.)

[lemon24]

Another one(?) – 15 https://scattered-thoughts.net/atom.xml entries from the last ~2 years just appeared in my feed; these may be duplicates (if so, entry_dedupe should copy the old recent sort key, if it doesn't already).

Update: Nope, they were "new" entries (not duplicates); also, entry_dedupe already handles the recent sort key.

[lemon24]

http://yosefk.com/blog/feed switched feed generators a few days ago, and I got 7 8-year-old entries in recent. I believe the root cause is the same (new feed has more entries than the old one).

Update: I checked, and this is indeed the root cause.

[lemon24]

Another one, with hundreds of entries: https://blog.startifact.com/rss.xml