Another perspective for spine/batch distinction?

[blp]

The implementation of batches and spines in DBSP has this structure:

• A batch is an immutable collection of tuples.
• A spine is a mutable collection of batches.

I wonder whether there'd be value in considering the following approach:

• A batch is either an immutable collection of tuples, or an immutable collection of two (or more?) batches.
• A spine mutably refers to one batch.

In the latter, a "batch" refers to its overall merged content, and in the background the physical representation of the batch can be replaced by an equivalent, that is, merges can take place without disturbing the batch's identity.

I don't know whether this perspective is different from the existing one in an important way, but I personally felt like I had a shift in thinking when it came to mind.