Lambda should fail cleanly and throw an error if there is an unhandled error inside.

I use Lambda for a very complex data manipulation action. However, I found out that when the code inside Lambda encounters an unhandled error (null/undefined reference) it will return the error message as the result of the Lambda step.

IMO this is not a very intuitive thing to do and we should just cleanly 500 the API if Lambda has a runtime error.