This is a really thorough end-to-end build. The part that stood out to me was using Floci to emulate S3 and SNS locally so the event-driven flow could be tested without touching real cloud resources — that keeps the feedback loop fast and the cost at zero while still exercising the real trigger path. Splitting the infrastructure and application work into separate pipelines also makes sense; a Terraform or Ansible change failing should not block an app deploy, and vice versa. The lessons on container isolation and Docker socket permissions are the kind of detail that usually only shows up after hours of debugging, so documenting the fix alongside the setup is genuinely useful. Did you find the readiness and liveness probes caught real issues during rollout, or were they mostly there as a safety net?