As a product lead, how do you decide when to sunset a feature vs keep iterating on it? What is the right balance for an agile squad/product?

A feature once built should on regular intervals under the same discovery process a PM would do for a new feature. The business case should tally to continue to invest or maintain a feature. I go through the discovery process as though the feature doesn’t exist, if the results are closer to the feature that is built I continue, if the results are different it is an opportunity to do something new/modern in place of the existing feature. Sunsetting a feature has two phases - 1. Discovery process results in a feature that is no longer used/needed or has a new avatar calling to pivot the existing one. This could be due to technical debt based limitation needs as well.
2. The business case to support it doesn’t tally by carefully considering the customer churn if any.
One recent example where I ended up sun setting a feature was when partnering with another offering made more sense than continuing to invest in a feature that was a sink hole and demanded changing the DNA of my offering.

Adding Greg F to chime in as well. Really appreciate the great, insightful responses!

This is one of the hardest things to do for a product organization. No one (PMs, employees, leadership, customers, etc.) likes to do this. What often happens instead is that we let things slowly degrade rather than actively manage.

For us, the key to the decision above is to make it as fact-based as possible. Knowing such things as how many customers use a feature, how big the technical debt is, customer satisfaction, market alternatives, profitability, etc. are all key to making informed, focused decisions. In this way we can avoid some of the emotional challenges and instead focus on the facts which lead to a decision. That being said, it is always paramount for the PM to take the leading role in this decision. Too often the drivers for sunset come from other teams (e.g. too many calls, etc.) but those shouldnt drive the decision. They can however, inform the choices so that the PM makes the hard call.