Nice! Are you planning to run this on K8s eventually? I've seen the control plane costs add up fast if you're not careful with spot instances and rightsizing.
This is a great example of how unpredictable AWS billing can derail projects. I've seen this exact scenario play out at multiple companies - sudden 700%+ cost spikes that force engineering teams to become billing experts instead of building product.
A few learnings from teams I've worked with who faced similar issues:
1. RDS cost surprises often come from I/O charges that aren't obvious upfront. Moving to reserved instances helps, but doesn't solve sudden usage spikes.
2. The "build your own monitoring" approach works but has hidden costs - engineer time, maintenance, alert fatigue, etc.
3. Many teams find that automated cloud optimization platforms (like CloudExpat, Spot.io, or CAST AI) can reduce costs 60-90% without the engineering overhead.
For anyone spending $10k+/month on AWS/Azure/GCP, it's usually worth getting a free cost analysis. Even if you don't use a platform, they'll often reveal billing patterns you didn't know existed.
Your Beacon project looks solid for self-hosters though - great work turning a painful experience into something useful!
I remember reading this when it was published, and the milestone breakdown Mitchell describes was exactly what got us from a theoretical solution to a production platform.
That combined with launching an MVP first rather than the "complete" vision. Shipping early meant we avoided the trap of spending months perfecting features no one actually used.