Empty cron workflows recurse forever

[josephjclark]

When a cron workflow runs and runs with only itself as an input, the state slowly gets so big that it'll cause the worker to crash

See history

[josephjclark]

So we have two problems here:

1. How do we stop this happening going forward
2. How do we fix workflows which are in this state

[josephjclark]

Hang on, is this just a leaf nodes thing?

I have a feeling that a single node cron job will be fine here. It's when we're returning multiple leaves that things are getting messed up

[josephjclark]

Almost certainly a leaf nodes thing

To avoid cron jobs going beserk, let's ignore leaf nodes which return empty objects or falsy values.

That at least ensures workflows without any implementation will be safe. Workflows with multiple leaves and cron jobs will have to resolve the state themselves anyway to ensure that the cron workflow runs correctly.

So that's not too bad.

Just need to work out how to clean up production and staging workflows which are in this state right now

[josephjclark]

The actual issue should be fixed. Next we need to consider what to do about any of these crazy recursive cron jobs that are live in prod. I might need to ask for help with that - needs creative thinking