Here's a quote from the article:

1. A proposal is made, the group tries to reach consensus. Modifications may be made to the proposal. Consensus may be reached, or it may be decided that no consensus is necessary in this case – a few members of the group can implement the decision without bothering the others.

2. If, after reasonable attempts by all sides to accommodate each other have been made, no consensus is reached, a vote is taken using the agreed voting method. The facilitator will decide the right moment for the vote. The right moment can be the subject of debate, and people may raise objections, but if you have a facilitator who can be held accountable then they need to be trusted to make the final decision. Making a group decision on how to make a group decision when in the middle of a contentious meeting is not a viable option.

3. The group considers how those who lost the vote can be accommodated – this may have been dealt with already in the stage 1 discussions but it is often worth raising again after the vote.

4. The debate between differing views continues over the long term, and decisions can always be re-considered if appropriate.

In reply to @infinisil:matrix.org
Instead the article suggests to try to first reach consensus, but if it can't be reached, do a vote (e.g. 2/3 majority)

infinisil: Requesting to join the team:

• motivation: Unify the disparate ways package sets are defined, improve quality and maintainability of Nixpkgs, participate in large-scale improvements to Nixpkgs.
• github username: tomberek
• sponsored by Flox
• best times would be apx 1800-0200 UTC Wed-Sun

• Added an entry for tomberek to https://github.com/nixpkgs-architecture/team-log
• Added to the current team list: https://github.com/nixpkgs-architecture#current-team
• Invited to the github organization