On a typical write back cache, the cache line is written back to memory (or to a lower hierarchy) only when needed: either some other core wants it or it needs to be evicted to make space for other data. For a write mostly task, if you write to the same line again, it is likely to be still in the M state and data can be written to it immediately. Later, when the line is written back, the separate writes are effectively coalesced. But if some other core is concurrently reading from the head of the log, it can be suboptimal as the cacheline will keep bouncing between M and S state generating coherence traffic. This is still more scalable than two threads writing to the same cacheline of course.