We've got programmable timeout configuration in our HTTP client, and different API clients set different values for it based on what we've observed about API performance.

But we also have problems where our communication with an API gets cut off. This doesn't happen very often, but a single network glitch can cause a background script to crash and delay important work until cron runs it again. For this reason we should also add the ability to tell the HTTP client how many times it should try to make a request before giving up.

POST requests cannot be automatically retried since making the request multiple times may have unexpected side effects.