But you can imagine that this gets rid of the classic GCE - you have a negative bubble template, which is supposed prevent oversubtracting the bubbles by removing emission that is dimmer at 4.8 GeV than at 1.8 GeV. But then it is allowed to (and does!) pick up positive emission at 1.8 GeV. The GCE!

There are some checks in the paper. Most notably, also in Fig 14 when they use a Flat Fermi bubbles template, and still get something that peaks at 20 GeV (though has a huge negative flux at low energies). It's not clear to me from the paper whether flat FB also picks up low-E residuals, be default.

It is unclear to me how you can use the resulting residual to measure the spectrum of the new excess, since it's partially forced by the fitting procedure. You can see that in Fig 14 when you shift the data driven Bubble template to 1.5 GeV, the spectrum of the excess shifts to lower energies.

Then, the portions of the data that have negative residuals and positive residuals at this point are allowed to float independently in the future analysis - and by definition any new "GCE" component has to be 0 at the pivot energy of 4.8 GeV (from the sentence above).

Also, the "negative Fermi bubbles technique" in Sec. 3.1 is odd. (Note: I have not tried to reproduce the analysis). They basically build a Fermi bubbles model that directly fits the 4.8 GeV data, as they note: "Even if [the GCE] is present near this [energy], it will be absorbed by these templates"

I think it is worth being skeptical. Both papers that isolate the high-latitude GCE (e.g., Calore 2014), or study variations of the Fermi bubble templates (e.g. Hooper & Slatyer 2013, Fermi-LAT Collaboration 2017) have not found a similar signal, even though it should be bright in those datasets.

I don't think that matters though -- even if you assume the "classic GCE" is astrophysics -- you should still recover it as a signal in your model, which you would then have to subtract in order to find the "high-energy GCE" in the remaining residual.

I always thought that this plot is super important for understanding the Swedish Covid Response. Sweden has one of the lowest rates in the world of 25-34 year olds living with their parents. Sweden could afford relative freedom in school closings without those kids coming home to Grandma and Grandpa

You’re right, I should have put “discovered” in quotations. I mostly believe you would start writing down most of the correct gravitational corrections without knowing the reason.

This is one of my favorite physics alternate histories. You can imagine a world where rocket tech outpaces GR. Building GPS is obvious - and when they start getting wrong answers, the math for the corrections is not so hard. GR would end up being discovered by a data analyst rather than a theorist.