• Expose latency and per target stats globally
• default flag to true

Make Q2 ticket for:

• remove avgTimeSpentMs, log if number of targets exceeds the threshold.

Make backlog ticket for:

• remove toAdditiveStats
• Avoid the tdigest copy in Summary copy and copy assignment operators
• Avoid the loop in SourceOperator
• More tests
  • test unwind case
  • compare actual e2e latency in kafka.js to measured latency, Set this in broker and remove special casing — log.message.timestamp.type=LogAppendTime
• We might want to put the information in the checkpoint summary stats which survive stream processor modifications. We can make a follow up ticket for this and discuss with PMs on the desired behavior.