Scheduling is not just a timer

When people say "post scheduler," the mental model is usually a simple queue: you add a post, set a time, it goes out. In practice that's rarely enough.

You need to know what happened afterwards: did it publish or not, if not — why, and can it retry automatically. You need to handle platform constraints — every API has its own rate limits and format requirements.

We spent a few weeks studying how existing tools are built and came to a conclusion: most of them oversimplify the error model. Everything works great until the first failure, and then manual intervention is required.

A good scheduler should be able to recover on its own.