Flotilla launches self-hosted multi-agent orchestration stack

The sharp bet here is that multi-agent reliability is mostly an operations problem, not a model problem, and Flotilla treats it like infrastructure from day one.

• –Shared `MISSION_CONTROL` context and lessons ledgers target the biggest failure mode in agent workflows: session amnesia.
• –PocketBase as a single-binary backend lowers ops overhead for small teams that want always-on coordination without managed cloud lock-in.
• –Staggered heartbeat/dispatcher workflows create built-in peer review loops across different models, which can reduce silent failure from single-model self-validation.
• –Vault-first secret handling plus self-hosted deployment makes this more enterprise-friendly for privacy-sensitive teams than chat-tab-based workflows.