Implement the core logic of a command-line alerting tool for vmstat-style data. For this coding problem, write a function that receives the input stream as a list of strings. The first non-empty line is a whitespace-delimited header containing metric names such as us, sy, id, etc. Each following non-empty line is one sample collected once per second. Monitor the column named by metric_name. A sample row is abnormal if that metric's value is strictly greater than threshold. Maintain a sliding window containing only the most recent window_seconds samples. Whenever the number of abnormal samples in the current window transitions from less than or equal to max_exceed_count to greater than max_exceed_count, emit an alert containing all abnormal sample rows currently in that window, in chronological order. Return all emitted alerts. Each alert is a list of the original sample row strings. If no alert is ever triggered, return an empty list. Your solution should process the stream online and should not keep unbounded history.