I ran the McBroken data through the Spatial Equity Data tool. It found McDonalds are slightly overrepresented in white areas while locations with broken ice cream machines skew Black and low-income.
I wonder why you chose just that one specific stat to highlight.
The data does seem to suggest that in an area annotated as "white" you are more likely to have location and it is less likely to have broken machines, whereas in an are annotated as "black" the opposite is true.
Those are the only two categories showing up as extremal on both tests.
Without digging into the methodology it's hard to get much more out of it. It's also not clear hwy "Low-income" doesn't show up as blue/over-represented on that chart.
It's possible that there's a disproportionate number of machines in minority areas misconfigured.
It's possible that there's a disproportionate number of machines in minority areas that are old-style manual-clean.
It's possible that there's a disproportionate number of machines in minority areas disabled for some other reason (ingredient supply disruption? Demand not worth the cost of keeping the machine on all the time?).
It is, indeed, possible that there is correlation without causation.
But the real answer is "We don't know why, and huh, isn't that interesting."