To maintain higher ingest rates, you want your latest chunk, as well as all its associated indexes, to stay in memory, so that writes to the chunk and index updates merely update memory. (The write is still durable, as inserts are written to the WAL on disk before the database pages are updated.)
If your chunks are too large, then writes to even the latest chunk will start swapping to disk.
As a rule-of-thumb, we recommend that the latest chunks and all their indexes fit comfortably within the database's shared_buffers. You can check your chunk sizes via the chunk_relation_size_pretty SQL command.
