This came up briefly last week, and we also discussed in the IRC meeting today. Given the increased release cadence, I think we would benefit from moving from release-based deprecation to time-based. That is to say, instead of deprecating in release X and removing release X+1, we would deprecate in release X and remove no less than N later and at least M releases later.
General consensus in the IRC meeting was in favor, however I wanted to start a thread to discuss to gather additional feedback and if everyone is in agreement, then we can solidify the actual details on what the new deprecation policy should be. Cheers, Joshua