Yes. I meant the single thread as an analogy, if a block is found, other blocks are worthless. (more or less) Longest chain wins.
My line of though was, that currently the only way to scale with the traffic (and lowering the fees) is increasing the block size (which is hard as I learned from recent events), or reducing the complexity which is less secure (most likely more controversial)
Splitting the chain is effectively increasing the block size, but without the increased hashing and validation overhead.
The usage growth seems to be more of exponential rather than linear. Sooner or later the block size will need to be 4 mb then 40 mb, then what is the real limit? Otherwise waiting times and thus the fees will just grow rapidly. I don't think that it is desirable.
With splitting the ledger, the block size can remain 1-2 mb for long time, only new partitions needs to be added on a schedule. This would also make better use of the hashing capacity.
Cheers,
Tamas