creates a queue of 1024 items - when performing metrics updates in a loop, this queue will easily overflow, creating an unreliable metrics flow.

There are several ways to improve the situation:

• count dropped messages such that they can be discovered
• allow configuring the queue size for applications with more fine-grained needs
• don't post full metric on channel - instead, post a notification that a new value is available and send latest value to server at send time - in a burst, the current code will send the least recent value that is already on the queue and drop any newer values
• drop the oldest values in the channel when it's full, instead of not posting the latest