ShardRegion bottleneck

There is another thread on this topic here What are the possible bottlenecks of using Sharding?