The stoichometry comes out with 2:1 moles of lfp to diatomic oxygen which is significant

The combustibility you're referring to is a legal definition not a scientific one

There's plenty of proof in academic literature. FePO4 is quite stable because of the quirks in iron's valance up to about 500C. But the combination of of lithium skews the valance effects at high temperatures to start losing oxygen at 250C.

Please review the following literature for more information:

1. C. Delacourt, P. Poizot, J-.M. Tarascon, and C. Masquelier, Nat Mater., 4, 254 (2005).
2. J.L Dodd, R. Yazami, and B. Fultz, Electrochem. Solid-State Let., 9, A151 (2006).
3. G. Chen, .XSong, and T. J. Richardson, J. Electrochem. Soc,. 154, 4627 (2007).

Phosphate will decompose into phosphate ions and oxygen given enough energy. The energy of the P--O bond is greater than Co--O but ultimately means that LFP batteries are also self-oxidizing but less so than lithium cobalt oxide

Reference

1. Nature of PO Bonds in Phosphates Benjamin Gamoke, Diane Neff, and Jack Simons The Journal of Physical Chemistry A 2009 113 (19), 5677-5684

What's seemingly more impossible is for the blue line to decline. I assume all lines represent some sort of integral which means number of games played have become negative YoY

But it's not as funny if you made the graph accurate