Initial chat. A take home project. Two interviews over the course of the take home. In each one we discussed the work I had done offline then worked through an issue together. At the end of the first I was sent the updated requirements for the take home to complete before the second interview. Final behavioural interview

You get what appears to be an automated invitation to complete a coding exercise, they claim that it'll take about 25 minutes, which seems too optimistic, it took me about 2 hours to complete as I documented and tested the solution. It's highly ambiguous and there's zero opportunity to ask questions. After submitting the solution, I got a rejection with an ambiguous response "Attention to detail: there may be more data on the running server than your solution accounts for.". I specifically documented that this would likely be the case, however due to the ambiguity I didn't know what scale to expect. Should I expect there to be hundred records or a billion records? The solution for a billion records would be much different than a solution for a hundred records, therefore I chose the safe route. Apparently the safe route wasn't good enough. They need to disambiguate the task a bit to ensure no one gets "lucky" with the right solution, I could have easily made the right choice, but it also could have just as easily been the wrong one due to the ambiguity. This company appears to rely too much on automation to validate their own assumptions, even their own feedback on other response confirms that assumption.