Performance Optimization

Batch requests, debounce filters, and reuse sockets

• Prefer long-lived WebSocket connections.
• Coalesce filters and de-duplicate subscriptions.
• Exponential backoff and jitter for reconnects.

Keep connections healthy and efficient

• Use ping/pong heartbeats; detect half-open sockets quickly.
• Compress payloads where supported; avoid excessive JSON nesting.
• Shard subscriptions by timeline vs. DMs to avoid overfetching.

Exploit temporal locality and derived views

Use in-memory caches for hot timelines, and precomputed indexes for common queries. Invalidate by event id and author.

Throughput on APIs and workers

• Connection pooling and reuse; avoid per-request TLS handshakes.
• Zero-copy parsing where possible; stream large responses.
• Batch writes and use idempotent upserts keyed by event id.

Find bottlenecks, not guesses