[Bug]: Killing a running durable execution does not kill child contexts

[dreams-and-thoughts]

Expected Behavior

All child contexts started within the parent execution should stop when the parent execution is stopped.

Actual Behavior

The child contexts continue to run, despite the parent execution being stopped.

Steps to Reproduce

1. Start a child context process using: context.run_in_child_context(child_process)
2. Stop the lambda execution from the console.
3. The execution status goes to Stopped.
4. The child context stays in 'Started' status, and the process continues, confirmed by log output.

SDK Version

1.3.0

Python Version

3.14

Is this a regression?

No

Last Working Version

No response

Additional Context

We have a use case where a process (running in a durable lambda) can be 'terminated' by a user via an API call.

[yaythomas]

I assume by "terminated... via an API call" you mean to call the StopDurableExecution API.

This will stop the durable execution, but it does not stop the current Lambda invocation. Once the durable execution is stopped, it will not accept any more checkpoints (not even SUCCEEDED/FAILED, the operation will remain STARTED). Lambda does not have a way to stop currently running Lambda functions.

If your Lambda code is already running, let's say it's in a step or a child context doing some work, this will keep running until it hits the next checkpoint, at which point it will raise an invocation error and stop the current invocation. This is likely to be at the end of current block of code, once it tries to checkpoint the result of that code.

[yaythomas]

I'm closing this issue as the behaviour is working as intended, but I've created a docs improvement issue so we can document this behaviour better.