Split the 5000-lamport signature fee into inclusion fee + resource fee
Priority fee unchanged.
Low-resource usage wins — votes & oracle updates can pay less than today. Compute-heavy, zero-priority traffic pays more (see examples).
| Example | Total Fee Before | Total Fee After | Δ |
|---|---|---|---|
| Vote (w/ Compute Budget Instructions) | 0.000005000 | 0.000004383 | −12.3% |
| Zerofi oracle (66vm...mMAD) | 0.000007568 | 0.000006288 | −16.9% |
| DFlow swap (high prio) (V7X2...SGs8) | 0.002007500 | 0.002202720 | +9.7% |
| OKX swap (mid prio) (3cE8...8Udh) | 0.000135980 | 0.000545640 | +301% |
| Pump.fun (zero prio) (5oGd...fY2) | 0.000005000 | 0.000162410 | +3150% |
Uniform hikes hit market makers and high-volume low-resource senders hardest, and still don’t price compute / locks / loaded data. This ties burn to requested scheduler cost instead.
Same reasons as priority fees: fee known before execution, faster fee payer check, better UX, and incentives to request accurate compute budgets.
Replay cost hits every node, not just the leader. A leader share is a harder economic problem to model and risks second-order effects on validator incentives and market structure (e.g. favoring inefficient programs over efficient ones, favoring toxic swaps over maker updates). 100% burn keeps current validator incentives mostly unchanged.
Well-formed votes with ~3765 requested units are cheaper with a 2500 base fee +
resource_fee_rate of 0.5. Naive votes requesting ~54k units would be much pricier pre-Alpenglow
— operators should right-size requests (as the SIMD notes).