The latest updates on your projects. Learn more about Vercel for Git ↗︎

If a guild goes on an outage, you won't receive updates to their presence or whether they still remain in the guild. This code change introduces potentially invalid guild members, right?

Both invalid GuildMembers and missing GuildMembers is not ideal.
It seems like the one way to avoid this issue completly is by using the GUILD_PRESENCES intent for now.

The other question is when is the GUILD_CREATE event fired?
Is it just

• on inintial connect after Ready packet (for each Guild)
• when bot joins Guild
• when Guild becomes available again

Or will reconnecting for example also trigger it?
(Thus is there a point in not clearing the member cache if the guild was never marked as unavailable at least)

After full reconnect it will be triggered too, but then you also could have missed members leaving and presences being updated during the reconnect. So the same issue remains. The „fix“ for what you describe as problem here would be to fetch all members in guildCreate event and have a ready event that doesn’t run once but with on on every ready that does that for guilds you need it in. There’s no way for discord.js to validate the cache against what exists on discord other than replacing it completely in those cases.

The „fix“ for what you describe as problem here

Yes, a "fix" as there is normally nothing else that will re-fetch the guild members, leaving the cache in an invalid state, without any obvious signs/warnings. This can lead to very weird, random seeming and even frustrating behaviour.

Missed presence updates are also not a worry since this problem only occurs when there is no GUILD_PRESENCES intent, in which case presence updates will not be received anyway.

And now you introduced a rather breaking change by fetching all members there. Which discord.js never does, it‘s upon the library user to decide whether they want to fetch all members or not. As already stated, the premise of your PR is wrong. If you need all members fetched you can do so yourself by listening to the respective event and fetching them again on a reconnect. It’s not the job of the library to do that for you.

Your right, what I did here does not make much sense.

If I am seeing this right though the GUILD_CREATE handler

let guild = client.guilds.cache.get(data.id);
  if (guild) {
    if (!guild.available && !data.unavailable) {
      // A newly available guild
      guild._patch(data);
    }
  } else {
    // A new guild
    data.shardId = shard.id;
    guild = client.guilds._add(data);
    if (client.ws.status === Status.Ready) {
      /**
       * Emitted whenever the client joins a guild.
       * @event Client#guildCreate
       * @param {Guild} guild The created guild
       */
      client.emit(Events.GuildCreate, guild);
    }
  }

will not update the guild if a full reconnect occurs and the GUILD_CREATE's are sent again, since the guild would never have been marked as unavailable?

Let's say someone did want to keep the members fetched.
I can't find an event that would tigger when a guild becomes available again (after being marked as unavailable) or for when a shard reconnected (not resumed) again successfully.