Some EC2 instances inside Elastic Beanstalk lose connection to db

we've got an Elastic Beanstalk app with 2-10 servers. It's got its own VPC and inside that VPC is an EC2 Instance with a MongoDB.

The MongoDB EC2 Instance is in the same data center and the same subnet as the EB app and has it's port configured so that the EB's Security Group can access it. We're in transition to MongoDB, so we started to just insert a few entries into MongoDB every now and then. This instance runs mostly runs under 1% of it's capacity.

But a few of the EB's instance lose connection on a regular bases. Not all of them. Only like one or two at a time.

I don't see us hitting any limits on Amazon's VPC limits page or on any of the EC2 instance. And the greater part of these instances does not seem to have connectivity issues, still I get connection timeouts frequently.

What am I missing? Is there anything else to look for? How come this only happens to some of the instances, while MongoDB is idling?