The concept of a “round” of auditing was introduced in RLATool to simplify some aspects of managing, conceptualizing and reporting on audits. But the current tool does not allow jurisdictions to do a bit more auditing “just in case” a contest they are part of needs to escalate beyond the current sample size. That can be a huge inconvenience if it forces teams to pack up, leave, and come back another day, after other all other jurisdictions are done with the previous round, just to audit a couple of extra ballots. It can also slow down the overall audit and cause deadlines to be missed unnecessarily.
I think Arlo should resolve this problem. One approach would be to allow jurisdictions to do extra auditing if they wish, and provide up-to-date input on how much extra auditing might be needed. Of course the tool needs to properly track audited votes to ensure that all calculations are based on a uniform random sample.
Alternatively, the whole concept of rounds could be made more flexible in other ways.
There are other reasons jurisdictions might want to do some extra auditing, e.g. if they want to investigate discrepancies found in an opportunistically audited contest.