Okay that makes sense, when we have very low throughput it looks like consistency wasn't an issue but at that rate (especially if it occurs in bursts) I can see

Okay that makes sense, when we have very low throughput it looks like consistency wasn't an issue but at that rate (especially if it occurs in bursts) I can see how the prior S3 implementation could fail to scale fast enough