Weird Alert Behaviors - firing same(and wrong) values on count() conditions

Hi,
I have set my alert using count() conditions, assuming an alert is fired when count of the metric hits over 100. My original “query A” simply counts for a matching query(counting occurance of slow logs in my case).

I went through 2 problems ;

  1. The initial alert was sent even though the alert condition wasn’t met.
  2. Same alerting and no-data notifications were kept firing which didn’t match real data. When I see Test Rule results, the alert value always seems to be fixed as ‘1201’.

This was my initial alert settings:

repeating alerts and no data notifications:

I don’t understand where this 1201 value is created…
After changing my alert setting using max() which correctly aims to my goal, an OK status was sent and false alerts stopped firing.
Any hints?

(Dec 12 updated)
changing to max() condition did not work. I am still seeing fixed metric values when I hit Test Rule.
Are these values cached in some way?