I find "per container per bean" difficult to apprehend. I would say "one bean per bean id in a container".Lets have an example to understand it. We have a bean class Sample. I have defined two beans from this class in bean definition, like:
<bean id="id1" class="com.example.Sample" scope="singleton">
<property name="name" value="James Bond 001"/>
</bean>
<bean id="id7" class="com.example.Sample" scope="singleton">
<property name="name" value="James Bond 007"/>
</bean>
So when ever I try to get the bean with id "id1",the spring container will create one bean, cache it and return same bean where ever refered with id1. If I try to get it with id7, another bean will be created from Sample class, same will be cached and returned each time you referred that with id7.
This is unlikely with Singleton pattern. In Singlton pattern one object per class loader is created always. However in Spring, making the scope as Singleton does not restrict the container from creating many instances from that class. It just restricts new object creation for the same ID again, returning previously created object when an object is requested for the same id. Reference