whole-tenant MIGRATION — process density (Corridor B, France→Huananzhi)
The migration itself as a process: the
macro Fokker-Planck of phase durations over
263 real cycles, and the
micro view of a single move
as seen from inside both clusters
(durable rows + PVC + Running pods sampled read-only).
NO_NEGOTIATE = mechanism-only — this maps the
MOVEMENT mechanism, not a contract-governed operation (honest).
← p(x,t) process explorer1 · macro — distribution over runs of each phase duration. Honest finding: the durable
\copy is near-instant (row-count is a STEP), so the move's real duration+variance live in the
KILL (whole-tenant death, med 338 s, heavy tail = Strimzi finalizer) and resurrect phases.
2 · micro — one real move from the clusters: Huananzhi dest fills 0→~83 k during replicate,
France source goes unreachable (real ns death), reverse leg, gaps=0.
3 · p(rows,t) ensemble — the within-move density over N instrumented runs. Honest: durable
row-count is a two-state STEP (0 or full), so this density is degenerate by design — it proves the copy is
near-instant, it does not "flow".
4 · death→resurrect dynamics — the genuinely-FLOWING part of the move: France Running
pods drop 5→0 at KILL then climb back at resurrect; the spread in bring-up time across runs = the kill_s
heavy-tail variance made visible. THIS is where the move's real dynamics live.