Hello, the documentation on https://docs.railway.com/reference/deployments#deployment-states and https://docs.railway.com/reference/deployments#singleton-deploys is not clear to me.

Here's how I understand it, please correct any inaccurate statements:
Let d1 and d2 be the old and new deployments of a service. The deployment goes as follows.

• Assuming healthchecks are st up

• d2 is spun up and healthy

• All new incoming traffic is now routed to d2

• d1 will live for OVERLAP_SECONDS, but will never receive any new traffic again assuming all goes well

• After OVERLAP_SECONDS, SIGTERM is sent to d1 and it now has further DRAINING_SECONDS for graceful shutdown.

• If d1 wants to invoke itself (e.g. if a cron job is triggered) during OVERLAP_SECONDS or DRAINING_SECONDS, it can do so with a HTTP request to the public URL and the request will already be routed to d2, ensuring continuity

Why I need the clarification is that I have a bunch of runtime cron jobs, one of them can take up to 2 minutes to complete. I don't want it to be killed mid-job upon new deployment. Configuring a large-enough OVERLAP_SECONDS should solve this, since the cron job perfoms the actual work by calling an endpoint using the service's public URL (my last bullet point).