See #61: For parallel_tests, it is neccessary to set the command name to something random like this SimpleCov.command_name "RSpec #{rand(100000)}" so parallel runs do not overwrite each other.

This is not obvious to users of the library and should be handled by SimpleCov automatically.

As a bonus, it would be great if instead of random numbers, some replicable ID would be available for each parallel test run, for example using a MD5 of the invoked shell command so that reported line hits do not add up along the way when the same test runs twice.