[UI/Metric improvement] Avoid realtime segment transient "bad" status

During real time segment transition (probably finishing one segment and creating a new one), we will often see "bad" and "red" status for the real time table. I am assuming it is transient because the status changes back to "green" shortly.

If the transient status is expected, we may change the indicator to be "yellow" or something else. Otherwise it is going to be scary that numbs alerting of true problems.

No files or tests are named. Start by reproducing a real-time segment transition and tracing how the real-time table derives its status; done means the transient transition no longer appears as a red/bad state, while genuine problems remain clearly red.

Written by the indexing model from the issue text.