Feedmaster queue pileup and the response time for ultra pipeline

Hi team,
I am getting huge variation in the response time while processing request in ultra pipeline(sample pipe not much logic or processing).
I have one ultra pipe running with multiple instances, and hitting 11000+ records to pipe from jmeter.
Most of the request 90% are getting response in less than 2 sec and 99% are at around 4 sec.
Avg time process is 1.2 sec but 1% request are taking more than 10 sec.

My usecase can not accept latency beyond 4 sec but i am facing issue here as some req. are taking over 10 sec. What is the reason and what should i do for this to make it consistent, please suggest ?
Avg time is ~1.2 sec so variation of 1 or 2 sec is still ok but i am experiencing too much variation of over 8 sec. Please check status below.

You may want to bring this up with Support. Without know all the details of how your pipelines are built, it would be difficult to troubleshoot this in the Community, nor should all those details be here.

Hi Dmiller,
There is only one pipeline main and it has just a mapper as pass through so nothing related to pipe design. I thought it may be needing some configuration etc so asked here instead of support, however i will reach to support on this, please let me know if you have any input as pipe design is not concern here.
Thank you!