Figure: a map of where KV-cache retention policy lives today, in four bands, organised by the decision each layer optimises rather than by any gap. The field did not leave a gap; it specialised, and the earlier "nobody has done this" framing is retired outright. Commercial products: Samsung CMM-D up to 512 GB per module and CMM-B up to 16 TB per box; the XConn Apollo 256-lane switch with MemVerge GISMO; Astera Labs Leo; Marvell Structera. Serving frameworks: vLLM with PagedAttention, the baseline every offload layer sits on; LMCache, now a first-class cross-tier KV layer with pinning, lookup, compression, batched movement, compute and I/O pipelining, cross-engine transfer and prefill-decode disaggregation; SGLang with RadixAttention; NVIDIA Dynamo. TraCT is a rack-scale, prefix-aware CXL KV cache implemented on Dynamo with direct GPU-to-CXL operations — CXL-native KV caching already ships, so this package is not the sole or first CXL-native KV-cache system. PNM-KV selects which KV pages are needed before paying the movement cost, from a processing-near-memory accelerator behind CXL. CacheWise predicts reuse from agent and tool-call state. Tutti removes the CPU from the GPU-to-NVMe KV I/O path. Related work this package builds on rather than competes with: H2O, SnapKV, Scissorhands, PyramidKV, Quest, Ada-KV, HeadKV, InfiniGen, LayerKV, KVPR and PagedAttention. What is genuinely open is not a missing implementation but a missing unification: whether one KV-State Control Plane — SELECT, PLACE, MOVE, ATTEND, OBSERVE — should decide relevance, reuse, residency and movement jointly instead of leaving each to a separate subsystem. Nobody, including this package, has built or tested that. One box inside it has evidence, and it is simulated and provisional: inside PLACE, letting identity-keyed policy state survive HBM eviction is worth plus 34.15 percentage points to LFU, from 15.56 to 49.71 percent, and plus 24.85 to EMA alpha 0.01, from 31.54 to 56.39 percent, at a 32 gibibyte budget, while EMA alpha 0.15 is unchanged at minus 0.21 percentage points — larger than the spread between policies, and testable with no CXL hardware. Retracted from this figure: per-head eviction over 640 GQA queues, since every head of every layer is required for every token and evicting a head produces wrong outputs rather than a cache miss, and the tracked unit is a (sequence, token block) of 16 tokens, 5.00 mebibytes, with per-head signals used only for scoring; alpha = 0.2, never validated, superseded by a simulated optimum near 0.05 with non-monotonic behaviour and no convergence onto LFU as alpha approaches zero; RoPE locality prefetch, since RoPE is a rotation applied to Q and K at compute time and carries no information about which blocks are needed next; and the 97 percent hit rate and 16 times user capacity that once framed the opportunity. The scoring function is not novel either: a decayed-frequency score is LRFU, Lee et al., ACM SIGMETRICS 1999, with alpha its documented LRU-to-LFU knob. No CXL hardware was measured. Simulation numbers are provisional, because a single data-structure defect in Revision 1 inverted four of five findings.

Evidence labels. Third-party capabilities are vendor specifications or the systems' own published claims, cited not reproduced. This package's persistence result is Simulated (provisional): a single data-structure defect in Revision 1 inverted four of five findings, so Revision 2 is provisional, not settled. Capacity arithmetic is an Analytical model and stands — 320 KiB/token, 2.50 GiB at 8K, 40.0 GiB at 128K, crossover with weights ≈ 427,000 tokens. LRFU attribution is External literature. Retracted from this figure: the 97% hit rate, 16× user capacity, α = 0.2, per-head eviction, RoPE-informed prefetch, and every "nobody has / the gap" framing. No CXL hardware was measured. Canonical numbers v6.0.