Perhaps someone can confirm my intuition:
A tree hash with leaves of size 1 (for example Blake2 in a tree-hash mode) would of course not be very efficient.
But it seems that it might be a suitable way (or maybe the only way?) to provide a verifier that's still useful if the underlying content is subject to intentional declared byte-range-elisions later. (That is: a form of digital redaction, where the original verifier still helps to assure that only the declared range has been altered.)
A content host performing elision would record the range, record the one or (usually) two internal tree node hashes that cover the to-be-elided range, then zero (or encrypt) the range.
When other parties want to verify the content is unaltered (except for the declared range), the host would provide the content, the declared range, and the up-to-two internal node hashes to be used in-place of the missing tree-hashed values. A root tree hash value would confirm that all other bytes in the file remain as when originally hashed.
Are there gotchas with this approach?
Any other better way to achieve the same end-result? (That is, use a compact-authentication value from when the content was whole, to still strongly verify unchanged ranges of content after some declared edit-range is removed?)