Is every alert or warning in my ad account actually costing me money?
No. Alert systems, both the ones built into ad platforms and third-party monitoring tools, commonly fire on statistical thresholds that don't account for sample size. A new campaign or a low-volume segment can trip a "cost per result increased" or "conversion rate dropped" alert purely from normal statistical noise on a handful of events, not from a real underlying problem.
Before reacting to an alert, check the volume behind the percentage. A conversion rate that "dropped 40%" sounds alarming, but if that's based on 5 conversions dropping to 3, it's within the range you'd expect from ordinary day-to-day variance. The same 40% drop on a segment converting hundreds of times a week is a different situation entirely and worth investigating.
As a rule of thumb, the smaller the sample behind a percentage, the less that percentage means on its own. Look at the raw counts first, then decide whether the alert reflects a real shift or just noise.