FA-58931 / Payroll withholding rules / Open access
Supplemental flat-rate withholding tiers: tier rounding stage · case 01
Crossing bonuses are withheld one cent more than the contract amount.
ROOT CAUSE
Each tier is rounded separately and then summed instead of rounding the combined withholding once.
VERIFIED REPAIR
Restore the contract rule at the tier rounding stage step: use `wh = (low * 22 + high * 37 + 50) // 100`.
Unsuccessful approach: The attempt still rounds per tier and truncates each, so the total is often a cent short.
Case contract
Input [ytd_supplemental, bonus] in cents. Supplemental wages up to a cumulative 1,000,000.00 are withheld at 22%; the part of cumulative supplemental wages above that is withheld at 37%. Withholding = (low*22 + high*37) rounded half-up to the cent once. Return [low, high, withholding].
Why this case matters
The mandatory 37% tier depends on cumulative supplemental wages and must split the crossing payment.
1 / The failure
Exit 1"""Failure Map reference implementation. Python standard library only."""
import json
N = 1
observations = []
def solve(x):
ytd, bonus = x
limit = 100000000
below_room = max(0, limit - ytd)
low = min(bonus, below_room)
high = bonus - low
wh = (low * 22 + 50) // 100 + (high * 37 + 50) // 100
return [low, high, wh]
def check(label, actual, expected):
observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
fixtures = [[('regression', [98797663, 40240022], [1202337, 39037685, 14708458]), ('regression', [82759503, 45428822], [17240497, 28188325, 14222590]), ('partial-repair probe (boundary)', [99999950, 101], [50, 51, 30]), ('partial-repair probe', [2185641, 20897159], [20897159, 0, 4597375]), ('boundary control', [100000000, 100], [0, 100, 37]), ('boundary control', [0, 0], [0, 0, 0]), ('normal control', [98022676, 574037], [574037, 0, 126288]), ('normal control', [3276820, 34567110], [34567110, 0, 7604764]), ('normal control', [49161621, 69310], [69310, 0, 15248]), ('normal control', [66640759, 96623], [96623, 0, 21257])], [('regression', [82759503, 45428822], [17240497, 28188325, 14222590]), ('regression', [99121299, 4951921], [878701, 4073220, 1700406]), ('partial-repair probe', [100000000, 6674805], [0, 6674805, 2469678]), ('partial-repair probe', [100000000, 2071189], [0, 2071189, 766340]), ('boundary control', [99000000, 1000000], [1000000, 0, 220000]), ('boundary control', [150000000, 33], [0, 33, 12]), ('normal control', [100000000, 52968], [0, 52968, 19598]), ('normal control', [2545996, 41743464], [41743464, 0, 9183562]), ('normal control', [100000000, 21687395], [0, 21687395, 8024336]), ('normal control', [374880, 70400], [70400, 0, 15488])], [('regression', [99121299, 4951921], [878701, 4073220, 1700406]), ('regression', [97165412, 46185905], [2834588, 43351317, 16663597]), ('partial-repair probe', [97846124, 8522], [8522, 0, 1875]), ('partial-repair probe', [100849097, 11351881], [0, 11351881, 4200196]), ('boundary control', [100000000, 100], [0, 100, 37]), ('boundary control', [0, 0], [0, 0, 0]), ('normal control', [688889, 1036750], [1036750, 0, 228085]), ('normal control', [99561826, 3108388], [438174, 2670214, 1084377]), ('normal control', [100000000, 75200], [0, 75200, 27824]), ('normal control', [98276481, 26306241], [1723519, 24582722, 9474781])], [('regression', [97165412, 46185905], [2834588, 43351317, 16663597]), ('regression', [99684298, 1209484], [315702, 893782, 400154]), ('partial-repair probe', [3045185, 81939], [81939, 0, 18027]), ('partial-repair probe', [107387654, 40335762], [0, 40335762, 14924232]), ('boundary control', [99000000, 1000000], [1000000, 0, 220000]), ('boundary control', [150000000, 33], [0, 33, 12]), ('normal control', [63981087, 6343691], [6343691, 0, 1395612]), ('normal control', [2334092, 64988], [64988, 0, 14297]), ('normal control', [105717987, 5743298], [0, 5743298, 2125020]), ('normal control', [27124697, 75911], [75911, 0, 16700])], [('regression', [99684298, 1209484], [315702, 893782, 400154]), ('regression', [98057948, 8504199], [1942052, 6562147, 2855246]), ('partial-repair probe', [63007595, 8068739], [8068739, 0, 1775123]), ('partial-repair probe', [64670743, 3021249], [3021249, 0, 664675]), ('boundary control', [100000000, 100], [0, 100, 37]), ('boundary control', [0, 0], [0, 0, 0]), ('normal control', [95783477, 7789713], [4216523, 3573190, 2249715]), ('normal control', [100000000, 76030], [0, 76030, 28131]), ('normal control', [100000000, 5769322], [0, 5769322, 2134649]), ('normal control', [19110078, 54047], [54047, 0, 11890])]]
for i, (label, args, expected) in enumerate(fixtures[N-1]):
check("%s %d" % (label, i), solve(args), expected)
print(json.dumps({"observations": observations, "passed": all(x["passed"] for x in observations)}, ensure_ascii=False))
raise SystemExit(0 if all(x["passed"] for x in observations) else 1)
| Boundary fixture | Actual | Expected | Outcome |
|---|---|---|---|
| regression 0 | [1202337, 39037685, 14708457] | [1202337, 39037685, 14708458] | Failed |
| regression 1 | [17240497, 28188325, 14222589] | [17240497, 28188325, 14222590] | Failed |
| partial-repair probe (boundary) 2 | [50, 51, 30] | [50, 51, 30] | Passed |
| partial-repair probe 3 | [20897159, 0, 4597375] | [20897159, 0, 4597375] | Passed |
| boundary control 4 | [0, 100, 37] | [0, 100, 37] | Passed |
| boundary control 5 | [0, 0, 0] | [0, 0, 0] | Passed |
| normal control 6 | [574037, 0, 126288] | [574037, 0, 126288] | Passed |
| normal control 7 | [34567110, 0, 7604764] | [34567110, 0, 7604764] | Passed |
| normal control 8 | [69310, 0, 15248] | [69310, 0, 15248] | Passed |
| normal control 9 | [96623, 0, 21257] | [96623, 0, 21257] | Passed |
SHA-256 / 392ddff0aa162ec1449fd2bf338638ce3c9981c8cc161f858d3e2346d6d18d89
2 / The unsuccessful fix
Exit 1"""Failure Map reference implementation. Python standard library only."""
import json
N = 1
observations = []
def solve(x):
ytd, bonus = x
limit = 100000000
below_room = max(0, limit - ytd)
low = min(bonus, below_room)
high = bonus - low
wh = low * 22 // 100 + high * 37 // 100
return [low, high, wh]
def check(label, actual, expected):
observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
fixtures = [[('regression', [98797663, 40240022], [1202337, 39037685, 14708458]), ('regression', [82759503, 45428822], [17240497, 28188325, 14222590]), ('partial-repair probe (boundary)', [99999950, 101], [50, 51, 30]), ('partial-repair probe', [2185641, 20897159], [20897159, 0, 4597375]), ('boundary control', [100000000, 100], [0, 100, 37]), ('boundary control', [0, 0], [0, 0, 0]), ('normal control', [98022676, 574037], [574037, 0, 126288]), ('normal control', [3276820, 34567110], [34567110, 0, 7604764]), ('normal control', [49161621, 69310], [69310, 0, 15248]), ('normal control', [66640759, 96623], [96623, 0, 21257])], [('regression', [82759503, 45428822], [17240497, 28188325, 14222590]), ('regression', [99121299, 4951921], [878701, 4073220, 1700406]), ('partial-repair probe', [100000000, 6674805], [0, 6674805, 2469678]), ('partial-repair probe', [100000000, 2071189], [0, 2071189, 766340]), ('boundary control', [99000000, 1000000], [1000000, 0, 220000]), ('boundary control', [150000000, 33], [0, 33, 12]), ('normal control', [100000000, 52968], [0, 52968, 19598]), ('normal control', [2545996, 41743464], [41743464, 0, 9183562]), ('normal control', [100000000, 21687395], [0, 21687395, 8024336]), ('normal control', [374880, 70400], [70400, 0, 15488])], [('regression', [99121299, 4951921], [878701, 4073220, 1700406]), ('regression', [97165412, 46185905], [2834588, 43351317, 16663597]), ('partial-repair probe', [97846124, 8522], [8522, 0, 1875]), ('partial-repair probe', [100849097, 11351881], [0, 11351881, 4200196]), ('boundary control', [100000000, 100], [0, 100, 37]), ('boundary control', [0, 0], [0, 0, 0]), ('normal control', [688889, 1036750], [1036750, 0, 228085]), ('normal control', [99561826, 3108388], [438174, 2670214, 1084377]), ('normal control', [100000000, 75200], [0, 75200, 27824]), ('normal control', [98276481, 26306241], [1723519, 24582722, 9474781])], [('regression', [97165412, 46185905], [2834588, 43351317, 16663597]), ('regression', [99684298, 1209484], [315702, 893782, 400154]), ('partial-repair probe', [3045185, 81939], [81939, 0, 18027]), ('partial-repair probe', [107387654, 40335762], [0, 40335762, 14924232]), ('boundary control', [99000000, 1000000], [1000000, 0, 220000]), ('boundary control', [150000000, 33], [0, 33, 12]), ('normal control', [63981087, 6343691], [6343691, 0, 1395612]), ('normal control', [2334092, 64988], [64988, 0, 14297]), ('normal control', [105717987, 5743298], [0, 5743298, 2125020]), ('normal control', [27124697, 75911], [75911, 0, 16700])], [('regression', [99684298, 1209484], [315702, 893782, 400154]), ('regression', [98057948, 8504199], [1942052, 6562147, 2855246]), ('partial-repair probe', [63007595, 8068739], [8068739, 0, 1775123]), ('partial-repair probe', [64670743, 3021249], [3021249, 0, 664675]), ('boundary control', [100000000, 100], [0, 100, 37]), ('boundary control', [0, 0], [0, 0, 0]), ('normal control', [95783477, 7789713], [4216523, 3573190, 2249715]), ('normal control', [100000000, 76030], [0, 76030, 28131]), ('normal control', [100000000, 5769322], [0, 5769322, 2134649]), ('normal control', [19110078, 54047], [54047, 0, 11890])]]
for i, (label, args, expected) in enumerate(fixtures[N-1]):
check("%s %d" % (label, i), solve(args), expected)
print(json.dumps({"observations": observations, "passed": all(x["passed"] for x in observations)}, ensure_ascii=False))
raise SystemExit(0 if all(x["passed"] for x in observations) else 1)
| Boundary fixture | Actual | Expected | Outcome |
|---|---|---|---|
| regression 0 | [1202337, 39037685, 14708457] | [1202337, 39037685, 14708458] | Failed |
| regression 1 | [17240497, 28188325, 14222589] | [17240497, 28188325, 14222590] | Failed |
| partial-repair probe (boundary) 2 | [50, 51, 29] | [50, 51, 30] | Failed |
| partial-repair probe 3 | [20897159, 0, 4597374] | [20897159, 0, 4597375] | Failed |
| boundary control 4 | [0, 100, 37] | [0, 100, 37] | Passed |
| boundary control 5 | [0, 0, 0] | [0, 0, 0] | Passed |
| normal control 6 | [574037, 0, 126288] | [574037, 0, 126288] | Passed |
| normal control 7 | [34567110, 0, 7604764] | [34567110, 0, 7604764] | Passed |
| normal control 8 | [69310, 0, 15248] | [69310, 0, 15248] | Passed |
| normal control 9 | [96623, 0, 21257] | [96623, 0, 21257] | Passed |
SHA-256 / e266c1a915233b571086ff0ff037b45884d609f3795cad6c844af750119c9123
3 / The verified repair
Exit 0"""Failure Map reference implementation. Python standard library only."""
import json
N = 1
observations = []
def solve(x):
ytd, bonus = x
limit = 100000000
below_room = max(0, limit - ytd)
low = min(bonus, below_room)
high = bonus - low
wh = (low * 22 + high * 37 + 50) // 100
return [low, high, wh]
def check(label, actual, expected):
observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
fixtures = [[('regression', [98797663, 40240022], [1202337, 39037685, 14708458]), ('regression', [82759503, 45428822], [17240497, 28188325, 14222590]), ('partial-repair probe (boundary)', [99999950, 101], [50, 51, 30]), ('partial-repair probe', [2185641, 20897159], [20897159, 0, 4597375]), ('boundary control', [100000000, 100], [0, 100, 37]), ('boundary control', [0, 0], [0, 0, 0]), ('normal control', [98022676, 574037], [574037, 0, 126288]), ('normal control', [3276820, 34567110], [34567110, 0, 7604764]), ('normal control', [49161621, 69310], [69310, 0, 15248]), ('normal control', [66640759, 96623], [96623, 0, 21257])], [('regression', [82759503, 45428822], [17240497, 28188325, 14222590]), ('regression', [99121299, 4951921], [878701, 4073220, 1700406]), ('partial-repair probe', [100000000, 6674805], [0, 6674805, 2469678]), ('partial-repair probe', [100000000, 2071189], [0, 2071189, 766340]), ('boundary control', [99000000, 1000000], [1000000, 0, 220000]), ('boundary control', [150000000, 33], [0, 33, 12]), ('normal control', [100000000, 52968], [0, 52968, 19598]), ('normal control', [2545996, 41743464], [41743464, 0, 9183562]), ('normal control', [100000000, 21687395], [0, 21687395, 8024336]), ('normal control', [374880, 70400], [70400, 0, 15488])], [('regression', [99121299, 4951921], [878701, 4073220, 1700406]), ('regression', [97165412, 46185905], [2834588, 43351317, 16663597]), ('partial-repair probe', [97846124, 8522], [8522, 0, 1875]), ('partial-repair probe', [100849097, 11351881], [0, 11351881, 4200196]), ('boundary control', [100000000, 100], [0, 100, 37]), ('boundary control', [0, 0], [0, 0, 0]), ('normal control', [688889, 1036750], [1036750, 0, 228085]), ('normal control', [99561826, 3108388], [438174, 2670214, 1084377]), ('normal control', [100000000, 75200], [0, 75200, 27824]), ('normal control', [98276481, 26306241], [1723519, 24582722, 9474781])], [('regression', [97165412, 46185905], [2834588, 43351317, 16663597]), ('regression', [99684298, 1209484], [315702, 893782, 400154]), ('partial-repair probe', [3045185, 81939], [81939, 0, 18027]), ('partial-repair probe', [107387654, 40335762], [0, 40335762, 14924232]), ('boundary control', [99000000, 1000000], [1000000, 0, 220000]), ('boundary control', [150000000, 33], [0, 33, 12]), ('normal control', [63981087, 6343691], [6343691, 0, 1395612]), ('normal control', [2334092, 64988], [64988, 0, 14297]), ('normal control', [105717987, 5743298], [0, 5743298, 2125020]), ('normal control', [27124697, 75911], [75911, 0, 16700])], [('regression', [99684298, 1209484], [315702, 893782, 400154]), ('regression', [98057948, 8504199], [1942052, 6562147, 2855246]), ('partial-repair probe', [63007595, 8068739], [8068739, 0, 1775123]), ('partial-repair probe', [64670743, 3021249], [3021249, 0, 664675]), ('boundary control', [100000000, 100], [0, 100, 37]), ('boundary control', [0, 0], [0, 0, 0]), ('normal control', [95783477, 7789713], [4216523, 3573190, 2249715]), ('normal control', [100000000, 76030], [0, 76030, 28131]), ('normal control', [100000000, 5769322], [0, 5769322, 2134649]), ('normal control', [19110078, 54047], [54047, 0, 11890])]]
for i, (label, args, expected) in enumerate(fixtures[N-1]):
check("%s %d" % (label, i), solve(args), expected)
print(json.dumps({"observations": observations, "passed": all(x["passed"] for x in observations)}, ensure_ascii=False))
raise SystemExit(0 if all(x["passed"] for x in observations) else 1)
| Boundary fixture | Actual | Expected | Outcome |
|---|---|---|---|
| regression 0 | [1202337, 39037685, 14708458] | [1202337, 39037685, 14708458] | Passed |
| regression 1 | [17240497, 28188325, 14222590] | [17240497, 28188325, 14222590] | Passed |
| partial-repair probe (boundary) 2 | [50, 51, 30] | [50, 51, 30] | Passed |
| partial-repair probe 3 | [20897159, 0, 4597375] | [20897159, 0, 4597375] | Passed |
| boundary control 4 | [0, 100, 37] | [0, 100, 37] | Passed |
| boundary control 5 | [0, 0, 0] | [0, 0, 0] | Passed |
| normal control 6 | [574037, 0, 126288] | [574037, 0, 126288] | Passed |
| normal control 7 | [34567110, 0, 7604764] | [34567110, 0, 7604764] | Passed |
| normal control 8 | [69310, 0, 15248] | [69310, 0, 15248] | Passed |
| normal control 9 | [96623, 0, 21257] | [96623, 0, 21257] | Passed |
SHA-256 / 4ba7f63eb2feed0bae5e0ccd973fcc6c03308980c916e7a2191c17c1738b2574
Verification & scope
A deterministic teaching model of a stipulated payroll rule with toy thresholds and rates. It makes no claim of conformance to any tax authority, statute or jurisdiction and is not payroll software. This reproducer isolates one failure mechanism. Results cover the supplied fixtures. Variants within a family share a test contract and should remain grouped when constructing evaluation splits. Related mechanisms with a shared evaluation_group must also remain together; these controlled models are not independent production incidents.
Observations recorded using Python 3.12.14 at 2026-09-29T14:46:31.418257+00:00.
Case digest / c8403d3421f01108838c25af3fc864498c564ba950115bf712ceb1995255e3e1