Currently, the materializationLagMillis metric is exported as a counter, which only allows monitoring changes over time but conceals tail latencies.

To improve observability and help detect latency trends, we need to export this metric as a histogram instead. This will enable us to better understand the distribution of latencies across the system and identify tail behaviors.

Key benefits of this enhancement:

• Improved visibility into latency patterns when interacting with SLS.
• Ability to measure tail latencies for better system performance monitoring and troubleshooting.