I interpreted this as a batch write, e.g. "write these 10k entries and then come back". The benefit of that is way less overhead versus 10k concurrent background routines each writing individual rows to the DB. The downside is, as you've noted, that you can't "stream" new writes in as older ones finish.

There's a tradeoff here between batch size and concurrency, but perhaps they've already benchmarked it and "single-threaded" batches of 10k writes performed best.