Date: 2026-11-03. State: Proposed.
Context#
The conversation mentions TimescaleDB, Redis and NATS, but volume, market and budget are not defined.
Decision#
Propose conventional PostgreSQL and outbox first; add TimescaleDB, Redis or NATS only on demonstrated need and validated support. Keep versioned event contracts.
Alternatives#
Operate all of them from day one; depend on an unverified extension; events only in memory. They add cost or loss of durability.
Consequences#
Initial polling adds latency and requires limits; performance will be measured. Tick/book data or high fan-out may change the decision.
Review trigger#
Measured throughput, retention, latency or consumers justifying a service/extension and an approved budget.