The code block action in a DTL is for writing arbitrary ObjectScript, not Javascript. It's commonly used for for data manipulation that can't be satisfied by the methods available in the FunctionSet; for example, extracting and decoding a base64-encoded PDF from an OBX:5.5 field and writing it to a file. It can also be used to interact with globals to maintain state between invocations of the DTL, or perform a database lookup, or even write values to the default device that will display in the Test tool. Very useful for debugging.

I would not recommend using it for operations that could potentially block. There's no built-in mechanism for setting a timeout so use a BPL for those cases.

I don't have a script for you, but if you'd like to obtain errors encountered with parsing messages against the identified DocType/Category, feed them through a service that assigns the expected Message Schema Category and send them to a business process that calls the Validate() method with the desired flags. It will return a %Status; you can extract its error strings with $System.Status.GetErrorText().

Do you need to embed the PDF file in the HL7 message so that it can be extracted as a PDF by the receiving system, or do you need to extract patient demographic and clinical or financial data from the PDF file and send the discrete data as field content in the HL7 message?