Choosing Between GUILD_MEMBERS Intent Caching and API Fetching for Large-Scale Discord Bots

Answer

For a bot that must stay under a 1.5 GB Node.js heap while resolving members in guilds ranging from 50 to 150 k, the GUILD_MEMBERS intent should be disabled unless the total unique member count across all guilds is known to be below roughly 800 k–1 M.

Confirmed facts

• In discord.js v14 each cached GuildMember object occupies about 0.8-1.2 KB of V8 heap (including user, presence and voice data).
• With a 1.5 GB heap limit, the cache alone can hold roughly 1.2-1.8 million members before exceeding the limit; practical pressure appears around 800 k-1 M total members.
• Calling guild.members.cache.sweep() removes entries from the cache, but if the GUILD_MEMBERS intent remains enabled the library will refetch any swept member on the next access, rebuilding the cache quickly.
• A sweep interval longer than the average member-lookup frequency (e.g., ≥5 minutes) yields net memory savings; frequent sweeps (≤30 seconds) erase the benefit.
• When the intent is disabled, each member lookup triggers a REST request. Staying safely under Discord's global limit of 50 requests/second is achievable by batching fetches (e.g., guild.members.fetch({ limit: 100 })) and rate-limiting the lookup queue to ~30-40 concurrent requests per second.
• Pre-warming only high-priority guilds is possible: keep the intent disabled globally, then on startup or when a guild becomes available call guild.members.fetch({ limit: 0, force: true }) for those guilds. This populates the cache for the selected guilds while leaving others to rely on on-demand fetches.

Likely explanation & recommendation

If the bot serves hundreds of guilds with an average size in the tens of thousands, the total member count will likely exceed the 800 k-1 M threshold, making the GUILD_MEMBERS cache consume most or all of the 1.5 GB heap. In that scenario:

1. Disable the GUILD_MEMBERS intent in the client options.
2. Identify a set of high-priority guilds (e.g., those where moderation or role-change frequency is highest).
3. On client ready or when a guild becomes available, pre-fetch members for those guilds with guild.members.fetch({ limit: 0, force: true }).
4. For all other member lookups, use batched guild.members.fetch() calls wrapped in a rate-limiter that targets ~30-40 requests/second.
5. Monitor process.memoryUsage().heapUsed and the rate-limiter's queue length to ensure staying below the heap limit and avoiding 429 responses.

If you can confirm that the total unique member count across all guilds is safely under ~800 k, you may enable the intent and optionally use a sweep interval of 5 minutes or longer to keep memory growth modest.

Missing diagnostic detail

To finalize the recommendation, please provide: What is the approximate total number of unique members the bot will encounter across all guilds? This figure determines whether the intent's memory footprint fits within the 1.5 GB budget.