Contents of this source
Remaining C1 gate after the counterexamples
Status: OPEN; the implications below are CONDITIONAL and are not new analytic estimates.
For one fixed allowed original background S and one actual strongly convergent row (u_j), a direct bound
e_S(P(u_j:j in F)) <= C(S,row) sum_{j in F} e_S(u_j)
for all nonempty finite F, with the constant independent of |F|, would still suffice. Bounded row energies give r(P_F)<=constant*|F|^(1/(q+1)), hence the frozen SR6 packing criterion, and therefore literal L1 continuity. The new two-entry counterexample does not refute such a bound with a larger constant.
In q=1, repeated use of the binary estimate with constant 2 gives a worst-case quadratic energy budget in |F|, hence linear root growth, which is insufficient. A DIRECT finite-family constant 2 would instead give square-root growth, which is sufficient. These statements must not be conflated.
Even this direct uniform finite-family bound is stronger than necessary: frozen SR8 requires only that every subsequence contain, for every eta>0, some finite nonempty F with r(P_F)<eta|F|. The reviewed p_* criterion is the uniform SR6 asymptotic rate, not an equivalence with SR8.
A second surviving route is ordered rooftop energy continuity for a fixed background (OG6), with estimates depending on that background and the fixed lower reference. The new modulus counterexample rules out uniformity over changing backgrounds using only energy and I, but leaves this fixed-background statement untouched.
Exact stopping point: no original fixed-S strongly convergent row with positive literal ED-data discontinuity has been constructed. No general SR8 finite minorant budget, fixed-S ordered-rooftop continuity estimate, or positive restriction-tail estimate has been proved. C1 remains OPEN.