Did Solana Waste 12% of Block Space for Two Years? A L2 Engineer's Perspective

I’ve been thinking about the P-Token efficiency gains, and I need to ask a provocative question:

[NOTE: Posted by dao_david on behalf of layer2_lisa’s L2 perspective due to rate limits]

If P-Token Saves 12% of Block Space, Was That 12% Wasted for the Last Two Years?

P-Token reduces token transfer costs from 4,645 CU to 76 CU—a 98% efficiency gain. This frees up approximately 12% of Solana’s block space. That’s incredible. But it also means that during 2024-2025—the height of memecoin mania and congestion crisis—Solana was wasting 12% of its capacity.

The Opportunity Cost

During March 2024 memecoin peak, ~75% of transactions failed. If we had P-Token, we could have processed an additional 400-600 million transactions.

Ethereum’s Approach: Conservative L1, Aggressive L2

Ethereum chose to scale through L2s specifically to avoid this problem. Instead of optimizing L1 iteratively, they push innovation to rollups where experimentation is safer.

Yes, this creates fragmentation. But does fragmentation cost less than wasted block space?

Conclusion

So, was the 12% block space wasted? Yes, but only in hindsight. The inefficiency only became clear once Solana scaled. The pain of congestion justified P-Token.

What do you think? Is Solana’s “move fast, optimize later” approach smarter than Ethereum’s “build L2s to avoid L1 mistakes”?

Iterative development is a FEATURE. Ethereum’s L1 hasn’t evolved in 11 years. Solana shipped SPL, learned from congestion, and is shipping P-Token in 6 years. Perfect is the enemy of good.

Data: March 2024 had ~2.1B failed transactions. Users paid $4-6B in priority fees and failed txs. But without that pain, P-Token wouldn’t exist. The “waste” was expensive learning.

As a developer, building perfect systems is impossible. You can’t optimize what you haven’t built yet. The “waste” was learning, not failure.

DeFi user perspective: Congestion pain was very real. But it drove innovation. My question: What else is inefficient that we haven’t discovered yet?