|Don't ask to ask, just ask|
If one had to produce the source filtering code from scratch, I might agree with you.
However, Filter::Simple offers the FILTER_ONLY interface, that makes it easy to specify code only transformation.
All this is going to be used for is getting people used to basic concepts like flow control, stack operations, and so forth.
One could pretty easily make the source filter so that it can be applied in chunks and each chunk could be incrementally removed:
The above could make list operations (push, pop, splice, etc) and flow control keywords return to their standard English form, while keeping file operations and other groups in Spanish.
If Filter::Simple works as advertised for simple code, then there is no technical reason not to use it. There may be pedagogical reasons not to use this approach. There are also similar reasons to use it. Which way will work better? I'd like to see some decent research that could answer that question.
Since the barriers to trying this with Filter::Simple are so low, I believe it is worth trying. If this is successful, many people around the world could find learing to program much easier with the help of Perl.