I gathered that Atlas alerts are design to retrigger once the condition is met. Then what does acknowledge mean in that case when providing any timeframe like forever? To me acknowledging an alert forever means that the alert would never to raise again.

In our case the number of connections for one cluster out of many goes slightly over the threshold, then shortly thereafter, goes below the threshold . The condition is 'flapping'. I know that this connection count condition is not leading to the number of connections to be exceeded. But, I don't want to ignore that alerts for all clusters, just the one in question.

The problem is that we will end of ignoring these alerts. We've done the work of examining the condition, plan to make the change, but have not done it yet. So, the retriggering of the alert for this cluster has no value added.

I think that either the UI ought to be changed to reflect how this behavior and not have an "Acknowledge for" option, which clearly doesn't work. Or the UI should respect that "Acknowledge for" period and not retrigger until that period has passed.