[valthon]

Automated underwriters waited whenever one account could not cover a request's entire outstanding bond. This left a 6 LIQSOL request unfunded even with five accounts each able to contribute 2 LIQSOL.

Contribute the minimum of the uncovered remainder, remaining exposure allowance, and wallet balance, rounded down to the contract's 0.01-token increment. Count each contribution against this pass's remaining funds and later passes' outstanding exposure. Keep outpost verification, oldest-first ordering, emergency stops, and settlement rules in place.

Regression coverage includes the five-underwriter example, exposure and wallet exhaustion across requests, previously bonded funds, caps lowered below existing exposure, and nine-decimal rounding around the minimum increment. The underwriter documentation describes shared coverage.

Validation:

• The shared-funding regression fails against the original planner.
• All 32 underwriting tests pass after the fix, compiled from this branch using cached native dependencies.
• The complete root Node tooling CI gate passes (coredump policy, dependency installation, and both generator registrations).
• Full Linux CI passes: ubuntu24, UBSan, ASan, assertions-on, and GCC build/test/package jobs, plus S1-S6 package verification.
• macOS arm64 CI passes: native build, parallelizable tests, and portable package verification.
• CI results are for commit bcbe0e73efefa6fb22deb717e6135ad820016c05.