My deployment is stuck using a cached Dockerfile that no longer exists in my repository. 

Current state:
- Repository has NO file named "Dockerfile" (renamed to Dockerfile.backup)
- Repository has nixpacks.toml configured for Python 3.12
- railway.toml has been deleted to force auto-detection

Problem:
Build logs still show "Using Detected Dockerfile" and are using Docker layers from weeks ago. The cached Dockerfile is missing recent Python files that exist in my current repository.

Build log evidence (timestamp 2025-12-18T19:06:03):
"Using Detected Dockerfile"
All build steps show identical timestamps, indicating cached layers

Expected behavior:
Should show "Using Nixpacks" and build with nixpacks.toml configuration.

What I've tried:
1. Added cache-busting ENV variable to Dockerfile (failed)
2. Changed railway.toml to builder = "NIXPACKS" (failed)
3. Renamed Dockerfile to Dockerfile.backup (failed)
4. Deleted railway.toml entirely (failed)
5. Multiple git commits and pushes (failed)

Request:
Please completely clear all build cache for this deployment and force a fresh build that will detect nixpacks.toml instead of using the phantom cached Dockerfile.

I'm not a developer, just using Claude. Everything was working super well, but every time I've had to try and change the Custom Start Command it wouldn't change. I would change it, save it, deploy and it would revert back to what it was before I changed it.