Decentralization → centralization seems to be a cycle. Why does it keep happening in DAOs?

Long before blockchains, there were decentralized-ish communities and alliances that tried to share power across many groups. A pattern shows up again and again in history:

When coordination gets expensive and threats get real, power concentrates.

One old example: the Delian League. It began as a coalition of Greek city-states cooperating for collective security. Over time, the alliance’s treasury, decision-making, and enforcement capability concentrated more and more — until it effectively became an empire run by the center.

I’m not bringing this up as an ideology point. I’m bringing it up because it feels like the same systems pattern DAOs run into:

The recurring “re-centralization” pressures

• Low participation: most members don’t vote → a small active group steers outcomes

• Decision bandwidth: too many proposals → people delegate or disengage → power concentrates in delegates

• Security/emergencies: you need fast action → emergency councils / pause keys appear → then they stick around

• Information asymmetry: a small group learns the system best → they become the de facto operators

• Coordination costs: the “center” becomes the only place that can move quickly and reliably

So decentralization often collapses not because people want centralization — but because the system rewards it under stress.

What I’m trying to understand (and build tests for)

If decentralization is a state you have to actively maintain, what mechanisms actually prevent drift back to central control?

Questions (real-world examples preferred):

1.  In production DAOs, what’s the #1 cause of re-centralization: apathy, delegation, emergency powers, or operator capture?

2.  Which mitigations actually worked long-term (not just on paper)?

• timelocks?

• role separation?

• rotation/expiry of special powers?

• caps on high-risk proposal throughput?

• stronger “version discipline” (preventing silent rule changes)?

3.  What’s the nastiest edge case where a DAO looked decentralized but wasn’t (soft capture)?

For context (not selling, just sharing the mechanism): I published a governance-only, hash-verifiable public package called DDD that tries to address drift via ruleset locking + patch-stack versioning + emergency ladders with time-bounds + appeals.

Canonical: https://github.com/Honest96-cyber/ddd-ruleset-2026-03-01/releases/tag/ddd-ruleset-2026-03-01

Mirror: https://drive.google.com/file/d/1IKoPLBhYm99uqwB-EsxlyaTzSlXw4f-W/view?usp=drivesdk

Verify: 00_Start_Here/MANIFEST.sha256.json inside the ZIP.

If you’ve got real examples (or “here’s how it failed”), I’d love to learn from them.