Spot on. Sounds like in the OP's case, people are putting a weight on the current velocity numbers and not looking at when features will be ready. Usually means the stories are too big.

Velocity over time is what you use to predict when future features will arrive. Yoyo velocity likely means the stories are too big or too small. If velocity drops and those features just move down the road. It should be clear as to why velocity drops (someone goes on vacation) or the story isn't broken down enough and takes too long.

Usually, in a velocity driven workflow, you don't need to lessen quality of code because developers point their stories and they should factor in testing into those points. If the stories are too big, break them into smaller stories so there is more accepted work each week. Like what we learned in math class... break a bigger problem into smaller problems...