Just corrected major bug by changing one assignment.
Purpose of including inline bubble #datavis of GISAID data was to demonstrate how major subvariants, not broken out by the CDC, were being obscured by CDC bundling those subvariants into ancestor for reporting purposes.
Which is what these #dataviz would have been doing, if only they'd been using the actual share of the given ancestor to set the size of its bubble. Instead, of, well (*clears throat with embarrassment*) the total sum of shares of the root variant and its descendants.
Would say that this was down to my focusing so much attention on dynamic legend and family grouping logic, but that ignores time spent adjusting how bubbles were aligned within tiles.
So, really, it was just that never bothered to consistency check the bubbles as drawn against data as given, assuming without validating that everything was correct.
Anyway, charts going forward now work as intended, reflecting when #CDC data hides more than it reveals.