Slow access from other EC2 servers after stop and start of EC2 instance

Our setup consists of a primary "server" c3.xlarge that includes SQL Server 2014 on Windows Server 2012 R2. There are client apps connecting to SQL Server from two other EC2 instances (t2.small) - let's call these "clients". Things worked fine until AWS sent a broadcast messages about underlying hardware degradation and mandated the "server" c3.xlarge be stopped and started, essentially moving it to a different physical host. From then on, performance of the client app became abysmal. Round trips to SQL Server to get data can take 500ms now where it used to take 10ms. SQL Server Profiler running on the "server" shows normal processing times, but we can see the StartTime of the next query in a sequence being 500ms from the EndTime of the one prior. To attempt to remedy the issue, one of the "client" instances was changed to "c5a.large", but it is still about 10x slower than before the "server" was stop-started.

Any thoughts? Is there a setting missing somewhere to put all 3 EC2 instances on the same physical host, if that is even possible?