Across the AST we have a lot of cases where we've made properties optional.
I think there were various reasons for why it was done this way (not that I can find the comments any more).

From the perspective of an AST consumer, undefined can create some very weird states. Two examples of this:

1. There are a number of boolean properties that are optional. On the surface this makes it look like the property with 3 meaningful states (true, false, or undefined), even though in reality it only ever actually represents 2 states.
   Usually in these cases the parser emits true or undefined (and never false), and in some rarer cases it emits true, and then uses false or undefined interchangeably depending on some internal, obfuscated criteria.
2. There are a few cases where an array property is optional. In these cases usually the parser will never emit an empty array, and will instead emit undefined.
   • From a memory-usage perspective, this could be justified (an empty array is really a heavy "empty" marker). But if that is our goal then these cases should be typed so that it is clear that there are never 0 elements (eg by defining [Foo, ...Foo[]])

We should standardise the AST to remove undefined/optional from all places, unless there is a very good reason for it to be optional.