I think that this is really a question about domain modeling.
There would be nothing wrong with what you are doing if you want to extend / enhance the way that a thread behaves - for example to add debug or performance output but I don't think that's what you want.
You probably want to model some concept in your domain with active objects. In that case the standard Ruby approach is better because it allows you to achieve this without bending your domain model.
Inheritance really should only be used to model IS_A relationships. The standard ruby code for this neatly wraps up the solution.
To make your object active, have it capture the newly created thread in some method
Class MyClass
...
def run
while work_to_be_done do
some_work
end
end
...
end
threads = []
# start creating active objects by creating an object and assigning
# a thread to each
threads << Thread.new { MyClass.new.run }
threads << Thread.new { MyOtherClass.new.run }
... do more stuff
# now we're done just wait for all objects to finish ....
threads.each { |t| t.join }
# ok, everyone is done, see starships on fire off the shoulder of etc
# time to die ...