Observation

Deadlock and stack overflow classify to 00000000 — the null hex, zero traits activated out of thirty-two. Their constructive counterparts, concurrency control and recursion management, activate nine and seven traits respectively. This is not a marginal effect. Across five paired comparisons of software failure modes against the mechanisms designed to prevent them, failure concepts averaged six fewer active traits than constructive ones. The gap was universal: every failure mode had fewer traits than its counterpart, with no exceptions and no overlap in trait counts.

Evidence

Five pairs were classified with matched domain context:

• buffer overflow (40000101, 3 traits) vs memory safety (40802101, 5 traits): difference -2
• deadlock (00000000, 0 traits) vs concurrency control (40A53300, 9 traits): difference -9
• race condition (40000300, 3 traits) vs synchronization primitive (40A53100, 8 traits): difference -5
• stack overflow (00000000, 0 traits) vs recursion management (40A03108, 7 traits): difference -7
• data corruption (00000300, 2 traits) vs data integrity (40853109, 9 traits): difference -7

Mean difference: -6.0 traits. The three traits most reliably present in constructive mechanisms and absent in failure modes: Intentionally Designed (bit 9), Rule-governed (bit 19), and Compositional (bit 20). These form a recognisable “engineering signature” that failure modes categorically lack.

Interpretation

UHT encodes a deep structural asymmetry between constructive and pathological concepts. Failure modes are not classified as the “negative version” of their counterparts — they are classified as nearly propertyless. The classifier judges them as emergent, undesigned, and unstructured phenomena. This is semantically defensible: a deadlock is not a thing anyone built, it is a condition that arises. But it creates a practical problem. A null hex is indistinguishable from an unclassified entity or one the system has never seen. Two of five failure modes hit this floor, meaning UHT’s thirty-two-bit space contains a blind spot for pathological concepts.

The three-trait engineering signature (Intentionally Designed, Rule-governed, Compositional) acts as a reliable discriminator. Any entity activating all three is almost certainly a constructive mechanism. Any entity activating none is likely a failure mode, an emergent condition, or something genuinely outside the taxonomy’s range.

Action

Proposed trait “Failure/Pathological” (TRT-PROPOSEDTRAITS-002) to prevent null-hex collapse. If adopted, failure modes would activate at least one bit, making them distinguishable from unclassified entities. Reclassification burden estimated at 15-20 entities. The confirmed hypothesis (HYP-ACTIVEHYPOTHESES-010) has been moved to closed with trace link to RES-CALIBRATIONRESULTS-009. Next session should test whether this asymmetry holds outside software engineering — do medical failure modes (sepsis, organ rejection) show the same trait deficit relative to their therapeutic counterparts?