The forking guidance in the driver is essentially copy-pasted from Mongoid, but this does not address use cases where the developer would use the driver directly instead of Mongoid. For example, in https://jira.mongodb.org/browse/RUBY-2167 the following use case is provided:

> A pretty straght forward and concise use case would be to use mongo for logging. A parent application might fork subprocesses, all of which communicate back to mongo to store their log messages.

When the driver is used directly, the recommendation should be to create new Client instances in each process instead of reusing and reconnecting existing Client instances as Mongoid applications tend to do.