This comes from the fact that the kernel does the accounting slightly differently. At a higher level it seems easier to just say I want 1G of memory and 1G of swap, but really to the kernel we're saying 1G of memory, and 2G of both together. This means that it's possible to get into a situation where your container is in fact only using 500M of memory and 1.5G of swap, which is more swap space than you configured for the container, but it doesn't mean the container will be able to exceed the 1G of RAM (or the total of 2Gs). The container cannot actually cause this to happen by itself. In order for this to happen there needs to be memory pressure from the outside, iow. from the host or other containers (or both).
If memory & swap were strictly separated then 1G of memory + 0 swap would effectively lock the container's used memory into physical RAM, which is usually not helpful, and if you add overprovisioning to this then it would become much easier to have the OOM-killer go on a rampage.
