I like your changes better. I will update mine to be the same as yours with additional comments and one more test to cover the NullReferenceException.

Interestingly, the NullReferenceException occurs only in beta5, due to #989.

@msftrncs Yes :)

I almost submitted #989 with a the break in it, but thought the EOS would always be there. I'm mentioning this for reference, as I should have added break when adding the while, because token is made null, and then without a break token is used in a reference. My gut said the break was needed, I just couldn't see why immediately. 😃

No worries at all. I didn't realize the stop early design for parsing NamedBlock until looking at this issue.
I looked into PowerShell code to see if process xxxx xxx, begin xxxx ... and end xxx ... should be fixed to generate tokens in a more expected/friendly way.
It doesn't look like a simple change to make it happen. The design is to stop parsing when process, begin and end don't end up with a valid NamedBlock, so tokenizing terminates in the middle as by design in those cases.
Fixing it in PSReadLine makes more sense.

@msftrncs Do you mind taking a look at my updates to #1049?
I took your changes and added additional comments and changed the tests a bit.
I didn't update your branch directly only because it's easier to work in my branch :)

Also, I invited you as a contributor to the PSReadLine repo, would you like to accept it? 😄

Close this one as #1049 has been merged. Thanks @msftrncs!