- "these two databases have different SQL dialects"
- "did we miss a few rows due to poor transaction-isolation when trying to query recently changed rows on the upstream database"
- "is there some checksum of a region of cells that accepts any arrangement of rows and columns that doesn't require me to think about ordering?"
- "is there a subtle data format conversion problem silently mangling my data?"
...I've toying with trying to find a way to serialize everything consistently into something that can be XOR'd, then compare the output of XOR for two tables in two different databases that should be identical, without having to do some giant order-by comparison. Ideally, if necessary, this could be done by eye with two query windows pointing at two different databases...
Basically, Datafold's datadiff, but in a way that could plausibly be home-rolled for on-premise applications (especially quick spot-checks), and not be a total maintenance nightmare...
https://github.com/datafold/data-diff
Don't have anything working yet, but it just seems like one could at least xor a bunch of integers and get something useful... Somehow.