Pubky does not require trusted caching servers. The source of truth is always the user's homeserver and public key domain (PKARR record). Caches, indexers, mirrors, CDNs, and search services are optional performance layers, not authorities. If a cache serves stale, filtered, or manipulated data, clients can fetch directly from the user's homeserver or use a different cache/indexer. The cache has no special power because it cannot change the user's identity, domain, or underlying data. In practice, Pubky treats caches similarly to web CDNs: useful for speed and scalability, but entirely replaceable. The user's public key remains the stable identifier, and the PKARR record determines where authoritative data is located. If a cache disappears, lies, censors, or goes offline, users retain a credible exit by switching to another cache, another indexer, or direct retrieval from the source. 🫳🎤