That is not odd at all.
You start off with a rough estimate based on the survival rates of other people with the same or similar diagnosis.
That's your first expected date. You start planning around it. (In her case, planning active measures to end her life while she could.)
Then you outlive it -- because after all, a median date implies people do outlive it. And that actually gives you hope and purpose which allows you less pain and perhaps even fewer symptoms.
And then a scan brings you right back to the median again, or much less.
Because the cancer isn't the diagnosis, or the scans. There are loads of other unseen variables.
This story repeats again and again; very few people die the way or when they expect to. A lot of people go suddenly after a long time of inexorable progress, some from secondary disease that can sometimes be avoided, and other people rally and then fade.