Where are you getting that figure? Everything I had ever read on the subject puts the half-life at more like 500 (five-hundred) years. While doing a quick double-check for this comment, it seems there was one sample with the unexpectedly long observed half-life of 15,000 years, still a couple orders of magnitude less than a million years. https://en.wikipedia.org/wiki/Ancient_DNA#Non-human_aDNA
I'm not an expert in this area but I do know a fair amount on the subject. I believe the section you linked is out of date.
The DNA we get back from old situations typically was preserved, by dryness/freezing, and it's still quite fragmentary. The actual "half life" of DNA is not that interesting- its the details of the DNA remnants that matter.
Yeah, you can still assemble a genome even if strands are broken. In fact this is done all the time as the size of the strands going into DNA sequencers are often only 100s or maybe 1000s of basepairs.
To be meaningful, half-life needs a Markov assumption, i.e., that the past and future are conditionally independent given the present. Or, more simply, at each instant, forget the past and for predicting the future use only the present.
Some interesting work in the many contexts where half-life works would be to say what the Markov property says about the mechanism of the decay or whatever are trying to predict in the context.
Not an animal, icecore, or that old, but ancient plants has been grown from seeds from permafrost.[0] So who knows what might be found and analyzed from all the icecores.
