[0.15.10] Duplicate favourites inconsistently

[Jon8RFC]

Sometimes servers get duplicated, but they disappear. Take a look in your own browser now to investigate before it goes away.

Shown here, I've removed my other favorites and this is just one server in my player-data.json list, but more shows up in the browser:
"SIdc9RwNnTVZB0ETKlkFn3OnDo+W9VTYczGUiRGX9WE="
http://imgur.com/a/f4h33


Two were displayed when I added one, and two went away when I removed one; this time three are showing. If I join and place an item, disconnect, join the other, the item remains and I spawn in the same place. Is this an issue with the browser showing duplicates somewhere? They aren't always duplicated, and they'll go away and occasionally show slightly different information, like the time being slightly different in this case, but in other instances the time will be the same. It's odd.

Again, I looked at player-data.json and there will just be the one entry, so I'm guessing something with the browser is showing a duplicate, or is there some chance that the server uses a poorly-designed host that mirrors data and doesn't disable them properly?

[Klonan]

Thanks for the report,

Does the issue with duplicate entries in the browse servers menu still occur?

[Jon8RFC]

I haven't noticed it in many weeks.

[Jon8RFC]

Uh oh, here's one instance of it again on 0.15.30 today. I haven't played since 0.15.26. I'm doubtful that something changed to correct and then break it, it just seems to happen without any rhyme or reason. The one marked as the favorite is the duplicate.

[Jon8RFC]

I spotted this again recently. I noticed that it occurred after the server's map was wiped and reset.

This was coincidence rather than the sole cause, but it has me wondering how Factorio validates a server's existence. It must take changed information from a map reset, loss of connectivity, unattended VM load balancing spinup/spindown, and count it as new and unique somehow, but doesn't have a way to purge the old listing because the IP and serverID are still valid.

Maybe a timeout purge of server update inactivity in some server list database needs to be shortened from tens of minutes of inactivity down to maybe 30 seconds? The servers get duplicated somehow, and servers that actually go down remain in the list as well, so there must be a purge setting that is too lenient.