That being said, the columns being converted from integers to floating point numbers is still an ongoing issue that precludes us from using the "combine datasets" feature reliably. I reached out to customer support yesterday and am awaiting a response in the meantime.
It turns out that the date issue is not a bug, but it is a gotcha that might confuse other people who deal with data stored in UTC time. TL;DR Dates don't have timestamps, so if you have date-level and timestamp-level data sources in Domo, they will not reconcile if your data isn't in the same timezone as your Domo…
I reached out to @DomoSupport within Buzz. Still an open question. They said that they're reaching out to the engineering team for why BIGINTs are being converted to DOUBLEs instead of LONGs (or why it's happening inconsistently between different tables!) The workaround for now is to run it through an ETL function.