This happends because you are using django.core.cache.backends.base.DEFAULT_TIMEOUT.
How we can fix it?

It seems to happen when cache-machine gets something from the cache, and then the same instance is re-used to cache something again. In that second time, somehow DEFAULT_TIMEOUT is not the original DEFAULT_TIMEOUT object, but a different one, so when the backend compares it, it finds it's different and happily accepts it as a normal timeout value, causing the failure later.

There is some code to guard against this, when it retrieves from the cache, it re-sets timeout to the correct DEFAULT_TIMEOUT, but somehow in that particular case it's broken, not sure why.