fix(pool): apply the sponsorship budget gate when counting candidates

[inakov]

Task 1 of the Rundler Production Task handoff doc. Metric-only: no change to which ops are eligible, returned to the assigner, or bundled.

Motivation

On rundler-robinhood-mainnet EP v0.7, rundler_op_pool_num_candidates read 317 while rundler_builder_assigner_ep_eligible_ops for the same entrypoint read 0. Every one of those ops was bundler-sponsored with a budget that no longer covered the current gas price, so the assigner rejected it on every pass. The pool's candidate check exempted sponsored ops from any price gate, so it counted them as bundleable indefinitely. That made the dashboard's num_ops_in_pool - num_candidates "stuck ops" panel read ~0 while essentially the whole pool was stuck.

Proposed Changes

• Replace the sponsored-op exemption in do_maintenance with the assigner's budget gate (op_meets_fee_requirements): a sponsored op is a candidate only while total_gas_limit * required_max_fee_per_gas <= max_cost. Non-sponsored ops keep the existing fee comparison.
• Route over-budget sponsored ops through set_underpriced() like non-sponsored underpriced ops. set_underpriced() only clears the time-to-mine tracker, so the op stays eligible and in best_operations. Side benefit: stranded sponsored ops no longer accrue days of time-to-mine that would skew the histogram if they eventually mine.
• do_maintenance returns the candidate count so tests can assert on it directly. The single caller ignores the value.
• Tests: sponsored op within budget is a candidate; one wei over budget is not, stays in the pool and in best_operations, and has its tracker cleared; a blocked op recovers when fees drop (re-evaluated each pass, not latched); non-sponsored ops unchanged.

Notes

• The pool uses state.uo_fees and the assigner uses its own required_fees. Both come from fee_estimator.required_op_fees(bundle_fees), sampled at slightly different times, so the two metrics should now track each other closely rather than exactly.
• The DA-gas / PVG eligibility path above this check still skips sponsored ops. Unchanged here, as the doc specifies.

Expected effect

num_candidates on robinhood EP v0.7 drops from ~317 to roughly the count of genuinely bundleable ops and moves together with ep_eligible_ops. The num_ops_in_pool - num_candidates dashboard panel becomes a usable stuck-ops signal.

Related: #1335 (Task 3, log spam). Task 2 (num_suspect_ops drift) will follow separately.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 63.68%. Comparing base (326e061) to head (b42ba8f).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

Files with missing lines	Coverage Δ
crates/pool/src/mempool/pool.rs	97.81% <100.00%> (+0.09%)	⬆️

... and 1 file with indirect coverage changes

Flag	Coverage Δ
unit-tests	63.68% <100.00%> (+0.08%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Components	Coverage Δ
rundler binary	10.26% <ø> (ø)
builder	76.67% <ø> (+<0.01%)	⬆️
dev	∅ <ø> (∅)
pool	67.04% <100.00%> (+0.27%)	⬆️
provider	30.90% <ø> (ø)
rpc	46.29% <ø> (ø)
sim	79.67% <ø> (ø)
tasks	∅ <ø> (∅)
types	71.72% <ø> (ø)
utils	53.40% <ø> (ø)

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.