| R027-01 |
0.2.6 historical evidence is mistaken for 0.2.7 execution |
false release claim |
separate SHA, version and evidence heads |
active record has no provenance class |
Owner |
| R027-02 |
registry v2 infers maturity from source presence |
overclaim |
require evidence and Owner status per combination |
orphan or duplicate public row |
Quality |
| R027-03 |
descriptor/preflight changes default behavior |
numerical/API regression |
additive path, negative tests and checkpoint replay |
default result changes |
Solver/Owner |
| R027-04 |
V&V harness refactor loses old evidence |
provenance loss |
leave old 188-module corpus intact and compare manifests |
missing historical artifact |
Quality |
| R027-05 |
C3D6/PENTA6 is not formulation-compatible |
false external validation |
preflight node order, kinematics and observables |
ambiguous mapping |
V&V |
| R027-06 |
WEDGE6 scope grows into mixed meshes/new physics |
schedule and validation risk |
static vertical slice and explicit exclusions |
unreviewed scope expansion |
Solver |
| R027-07 |
mesh quality reduced to aspect ratio |
invalid domain claim |
multidimensional diagnostics and Owner policy |
universal cutoff proposed |
V&V |
| R027-08 |
low-order locking is treated as a bug fix target |
unjustified formulation change |
diagnostic comparison and separate formulation study |
threshold weakened to pass |
Solver/Owner |
| R027-09 |
1M benchmark becomes a marketing claim |
misleading scaling |
hardware/topology/resource verdicts |
no resource context |
Performance |
| R027-10 |
stretch work blocks the bounded core |
release drift |
optional WP13 with independent decision |
research route enters core claim |
Owner |
| R027-11 |
full regression runs too frequently or too late |
wasted time or late regression |
T0-T3 policy and declared checkpoints |
no checkpoint after cross-cutting change |
Quality |
| R027-12 |
release metadata drifts from package/tag |
unusable artifact |
WP01 and WP14 exact version checks |
version mismatch |
Release |