Summary

Instances that were paired before v0.7.2 (when the project dropped the custom whatsmeow-lib fork and switched to official upstream go.mau.fi/whatsmeow) never receive the NCT salt, and therefore can never generate a <cstoken> for first-contact (cold) 1:1 sends. This makes error 463 (NackCallerReachoutTimelocked) permanent for any pre-v0.7.2 instance trying to message a number it has never talked to before — it never self-heals, no matter how many times the instance reconnects.

Root cause

• generateCsToken (whatsmeow cstoken.go) requires cli.Store.NCTSalt.GetNCTSalt(ctx) to return a non-empty salt. We confirmed via direct Postgres inspection that whatsmeow_nct_salt is completely empty (0 rows) for an instance that has been connected and working for months.
• The salt is only delivered through two whatsmeow-internal paths: (1) the one-time History Sync payload sent right after a fresh QR pairing (message.go, historySync.GetNctSalt()), or (2) an App State Sync patch (regular_high category, nct_salt_sync mutation index) fetched via FetchAppState.
• Instances paired before v0.7.2 went through their one-time History Sync using the old custom fork, which had no concept of NCT salt at all — that one-time opportunity is gone forever for those instances.
• Going forward, regular_high should self-heal via handleAppStateSyncKeyShare, but that function calls FetchAppState(ctx, name, false, onlyResyncIfNotSynced=true) — and onlyIfNotSynced skips categories that already have some stored version, even if that version predates cstoken support and never actually captured the salt mutation. So already-synced-but-stale categories are never re-fetched, and the instance is stuck permanently.

How Baileys solved the identical problem

Baileys hit the exact same gap and fixed it in WhiskeySockets/Baileys#2438 with a dedicated ensureNctSaltSynced() (in chats.ts): on every connection.update → open, if no salt is stored yet, it force-fetches regular_high on its own — independent of the normal "already synced" gating, guarded on the persisted regular_high version so it bootstraps at most once and never loops for accounts that legitimately have no salt yet.

Suggested fix

Port the equivalent of ensureNctSaltSynced() into evolution-go (or ideally into go.mau.fi/whatsmeow itself, so all consumers benefit): on events.Connected, if cli.Store.NCTSalt.GetNCTSalt(ctx) is empty, call cli.FetchAppState(ctx, appstate.WAPatchRegularHigh, true, false) once (forcing fullSync=true, onlyIfNotSynced=false specifically for this bootstrap case), guarded so it only runs once per process/instance to avoid loops for accounts that genuinely have no salt provisioned server-side yet.

Impact

Affects any instance paired before the org's whatsmeow upgrade (in our case, before 2026-07-03 / v0.7.2). For anyone doing bulk migrations of older instances (e.g. from a different provider) into evolution-go, this silently breaks first-contact/cold outreach sends indefinitely, with no error message pointing at the actual cause (surfaces only as server returned error 463 on /send/text, identical to the generic reach-out time-lock).

Environment

• evolution-go v0.7.2 (go.mau.fi/whatsmeow v0.0.0-20260630180629-b572e5bcb92b)
• Instance paired 2026-02-18 (pre-v0.7.2, custom whatsmeow-lib fork era)
• Confirmed via direct query: whatsmeow_nct_salt = 0 rows, whatsmeow_privacy_tokens = 0 rows for the affected contact, whatsmeow_lid_map correctly populated (so LID resolution is not the issue)