I recently discovered that invoking buffer() on a ConcatSource incurs a massive performance overhead by invoking Buffer.concat, particularly if there are layers of CachedSource -> ConcatSource -> CachedSource -> ConcatSource.
Invoking source() and the calling Buffer.from is considerably faster because it avoids the intermediary allocations.
We have writev nowadays, so there's no reason we couldn't keep the data as Buffer[] all the way until the final write to the file system and avoid repeated copying of memory during the concatenation layers.