I recently noticed that many of the buildings in San Diego County are being sourced from Microsoft ML buildings instead of from this higher-quality Esri dataset that covers pretty much all of San Diego County. I suspect that either this dataset isn't being consumed, or that there's an issue with the data conflation, leading to Microsoft's buildings being given a higher priority.

For context, I'm an OpenStreetMap mapper who primarily adds buildings from the aforementioned dataset. Rapid recently started sourcing the Esri Community Maps building footprints from OvertureMaps instead of from the many separate sources in ArcGIS, which is how I came across this issue.