I don't think the conjecture is correct. The content provided value to a person that read it. They rewarded the author with sats for this value. When the post is now removed, the value transfer already happened... The reader already got their value from the post. So it doesn't sound like returning the sats is a right approach.