Feature Gated: remove obsoleted FeeRateGovernor from fee calculation

[tao-stones]

Problem

• lamports_per_signature in blockhash_queue and nonce accounts is not used for fee calculation.
• Actual fees are fixed via bank.fee_structure.lamports_per_signature = 5_000.
• The only role of lamports_per_signature in the blockhash_queue is to enable test-mode zero fees when it's set to zero.
• Current behavior derives this value through fee_rate_governor, which is overly complex and introduces inconsistency, especially in early banks. (example)

Summary of Changes

• At genesis, store target_lamports_per_signature from genesis_config into the blockhash queue.
• For subsequent banks, copy lamports_per_signature from the parent bank.
• This keeps the intended behavior (e.g., zero-fee testing) and removes reliance on fee_rate_governor.
• Feature Gated, to avoid unintended fork divergence, this change is behind a feature gate (see Consensus divergence between v2.2 and master on Testnet #5454).

Note: impact on testnet

Genesis Config for three public clusters

Creation time: 2020-03-16T14:29:00+00:00
Cluster type: MainnetBeta
Genesis hash: 5eykt4UsFv8P8NJdTREpY1vzqKqZKvdpKuc147dw2N9d
...
FeeRateGovernor { lamports_per_signature: 0, target_lamports_per_signature: 10000, ... }

Creation time: 2020-02-04T16:35:32+00:00
Cluster type: Testnet
Genesis hash: 4uhcVJyU9pJkvQyS88uRDiswHXSCkY3zQawwpjk2NsNY
...
FeeRateGovernor { lamports_per_signature: 0, target_lamports_per_signature: 0, ... }

Creation time: 2020-08-10T17:36:56+00:00
Cluster type: Devnet
Genesis hash: EtWTRABZaYq6iMfeYKouRu166VU2xqa1wcaWoxPkrZBG
...
FeeRateGovernor { lamports_per_signature: 0, target_lamports_per_signature: 10000, ... }

Notice that testnet sets target_lamports_per_signature to 0 in its genesis config, but it is not a zero-fee environment today. (At some point, the lamports_per_signature value in the blockhash queue was updated to a non-zero value.)

With this PR enabled, a validator restarted from an existing snapshot will inherit the non-zero lamports_per_signature from its parent bank and continue charging normal fees.

However, if the validator were to restart from genesis, it would enter a zero-fee mode, since the genesis bank would initialize the blockhash queue with lamports_per_signature = 0.

Fixes #5782

[codecov-commenter]

Codecov Report

Attention: Patch coverage is 86.20690% with 8 lines in your changes missing coverage. Please review.

Project coverage is 83.4%. Comparing base (13799bc) to head (1025017).
Report is 2 commits behind head on master.

Additional details and impacted files

@@           Coverage Diff            @@
##           master    #5749    +/-   ##
========================================
  Coverage    83.4%    83.4%            
========================================
  Files         850      850            
  Lines      377790   377648   -142     
========================================
- Hits       315166   315049   -117     
+ Misses      62624    62599    -25     

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[brooksprumo]

Checking in, what's the status of this PR?

[tao-stones]

Checking in, what's the status of this PR?

PR is ready for review. Have run it in testnet for several days, looked OK. Not on the top of list atm, but appreciate reviews.

[tao-stones]

After rebase, it seems to break sbf stable test due to test program is not deployed. I'm taking it back to draft for investigation. Please hold on reviewing. Sorry about back-n-forth.

failed stable-sbf test

---- test_no_panic_rpc_client stdout ----
  | Waiting for fees to stabilize 1...
  |  
  | thread 'test_no_panic_rpc_client' panicked at tests/simulation.rs:86:10:
  | called `Result::unwrap()` on an `Err` value: Error { request: Some(SendTransaction), kind: RpcError(RpcResponseError { code: -32002, message: "Transaction simulation failed: Error processing Instruction 0: Unsupported program id", data: SendTransactionPreflightFailure(RpcSimulateTransactionResult { err: Some(InstructionError(0, UnsupportedProgramId)), logs: Some(["Program 11157t3sqMV725NVRLrVQbAu98Jjfk1uCKehJnXXQs invoke [1]", "Program is not deployed", "Program 11157t3sqMV725NVRLrVQbAu98Jjfk1uCKehJnXXQs failed: Unsupported program id"]), accounts: None, units_consumed: Some(0), return_data: None, inner_instructions: None, replacement_blockhash: None }) }) }
  |  

<br class="Apple-interchange-newline">

[tao-stones]

After rebase, it seems to break sbf stable test due to test program is not deployed. I'm taking it back to draft for investigation. Please hold on reviewing. Sorry about back-n-forth.

failed stable-sbf test

The failing test is flaky, pretty sure is a timing issue - test transaction could be sent before simulation program is deployed. This PR makes TestValidator starts up quicker by effective skip this step. Plan to address test's flakiness in own PR.

[tao-stones]

After rebase, it seems to break sbf stable test due to test program is not deployed. I'm taking it back to draft for investigation. Please hold on reviewing. Sorry about back-n-forth.
failed stable-sbf test

The failing test is flaky, pretty sure is a timing issue - test transaction could be sent before simulation program is deployed. This PR makes TestValidator starts up quicker by effective skip this step. Plan to address test's flakiness in own PR.

#6587 updated TestValidator that addressed the flaky downstream tests. Rebased to bringing this PR back for review.

@brooksprumo @apfitzge, for refreshment, it is part of #3303

[tao-stones]

To clarify what's expected behavior when this feature gate is activated:

This feature gate changes how lamports_per_signature is managed in the blockhash_queue.

1. Behavior After Feature Activation (Normal Operation)

• lamports_per_signature is copied from the parent bank into the blockhash_queue.
• The fee_rate_governor is no longer involved in computing or updating LPS.

Impact on fees: None.

• If LPS was 0, it remains 0.
• If LPS was non-zero, it will never become zero — this preserves current FeeRateGovernor behavior.

2. Validator Restart Scenarios

• Restart from Snapshot: the validator resumes by copying lamports_per_signature from the parent bank in the snapshot.
• Restart from Genesis: bank0's LPS is initialized from genesis_config.target_lamports_per_signature. This corrects two current inconsistencies:
  1. If target_lamports_per_signature != 0: bank0 will not be zero-fee (as it incorrectly is today).
  2. If target_lamports_per_signature == 0: the validator becomes a zero-fee cluster permanently, since LPS will remain 0 in all descendant banks.
     • This unlikely case could unintentionally turn Testnet into a permanent zero-fee environment if it restarts from current genesis.

I'll test it out by restarting this PR-built validator (with feature enabled hardcodedly) on three public clusters, make sure it's catchup, and stay up for an epoch.

(edited)