14.8 Consistency, Concurrency & Distributed State

1. What It Is in Plain Words

Consistency models define how and when updates to data become visible across all nodes and users in a distributed system. Concurrency control prevents race conditions when multiple users or processes attempt to modify the same resource simultaneously.

---

2. Consistency Models Applied in TruthGuard

1. Strong Consistency for Audit Custody & Profiles:
   • Updates to user permissions (Journalist -> Analyst) and evidence custody actions must be immediately consistent across all transactions. PostgreSQL transactions with ACID guarantees prevent dirty reads.
2. Eventual Consistency for External Fact-Checking:
   • Fact-checking syndications (Snopes, PolitiFact) update periodically. A 5-minute replication delay between external fact-checker releases and local cache population is fully acceptable for threat telemetry.

---

3. Concurrency Control & Race Condition Defenses

Optimistic Concurrency Control (OCC)

For multi-analyst case editing, TruthGuard employs optimistic locking using an updated_at timestamp column:

UPDATE public.incidents 
SET status = 'investigating', updated_at = now() 
WHERE id = :id AND updated_at = :last_seen_timestamp;

If another analyst modified the case record concurrently, the update count returns 0, prompting the user with a conflict resolution modal rather than silently overwriting evidence notes.

Distributed Locking for File Deduplication

When parallel ingest streams submit identical video files, a distributed Redis lock (SET lock:hash:<sha256> true NX EX 30) ensures only one worker initiates GPU processing, while sibling requests block until the lock releases.

---

4. 30-Second Spoken Interview Answer

"We apply a hybrid consistency model tailored to forensic data requirements: strong consistency in PostgreSQL for user authorization and custody audit trails, and eventual consistency for external fact-check feed synchronizations.

To prevent race conditions during concurrent case investigations, we use optimistic concurrency control with timestamp checks, paired with distributed Redis locks to prevent duplicate GPU processing on identical file hashes."