Kernel OOM killer prevents container processess from using swap to overcommit memory

[Pleune]

My system has 64GiB of ram, but only 30GiB freely availaible. The host is configured that processes will dig into swap. For example, my tail /dev/zero will run basically forever as zswap easily compresses the overflow. If I set a container limits.memory: 60GiB and limits.memory.swap: 'true', in the container tail /dev/zero gets killed by the host kernel OOM killer after using up the available 30GiB. If I change to limits.memory.swap: 100GiB, then my tail test is happy to use up swap space.

Is there a way to limit the instance to use only 60GiB of memory at all times, but allow using some swap if needed?