May 27, 2026
Spot It! but make it server drama
Raft Consensus with a Minority of Nodes
A nerdy math twist says broken networks might keep working — and commenters are losing it
TLDR: A new post argues computer clusters may keep working with less than half their machines active if the groups are chosen in a very clever math-based way. Commenters were split between "mind blown," "prove it better," and "pretty sure researchers already did this," turning the thread into a mini nerd cage match.
The internet’s distributed-systems crowd got a fresh brain scramble this week after Rohan Padhye proposed a gloriously "wacky" idea: what if a cluster of computers could keep agreeing on updates even when fewer than half of them are active? Normally, these systems play it safe by demanding a majority. Padhye’s twist borrows from the same math behind the card game Spot It!, where every pair of cards shares exactly one matching symbol. Translation for non-math people: he’s suggesting a clever way to arrange the rules so smaller groups can still overlap in just the right way and avoid chaos.
And the comments? Absolutely eating it up. One camp was delighted that someone dared to mess with a sacred cow, cheering that the post "flipped" their understanding of network failures on its head. Another group immediately turned into the internet’s unpaid fact-check squad: cool idea, they said, but show the proof more clearly and don’t forget the scary side of the story — if two isolated groups can both think they’re in charge, that’s how you get a nightmare split. Then came the classic "actually..." energy, with one commenter noting that researchers behind Flexible Paxos explored related ideas years ago.
So yes, the vibe was part awe, part nitpick-fest, part academic turf war. The funniest subplot is that a children’s card game has somehow become the main character in a serious argument about keeping databases alive. Peak internet.
Key Points
- •The article proposes a modified Raft design in which some minority subsets of nodes can still make progress if quorum membership is carefully constrained.
- •Standard Raft commits log entries and elects leaders using majorities, relying on the fact that any two majorities overlap in at least one node.
- •The article explains that this quorum overlap is the key safety property that preserves committed knowledge across elections and commits.
- •Spot It!/Dobble is presented as an example of a structure where any two sets intersect exactly once, using a deck built from finite projective plane mathematics.
- •The article connects finite projective plane intersection properties to the design of alternative quorum systems for consensus.