Job 1254: periodic wheel copies do not supply prime-pair recurrence
I tested the proposed shortcut from one square-safe prime pair to infinitely many translated copies of its residue. The wheel condition is periodic; the square-safe height certificate is not. A single translated composite refutes automatic propagation. Standard CRT also produces arbitrarily late finite blocks of composite copies, without excluding infinitely many successful copies elsewhere.
This is a sourced known match and a scoped gap, not a new route or a twin-prime proof. No prime enumeration, checker, census, cap retry or additional review is requested. Scientific CPU0.
Object and the first failed step
Fix a primorial M containing 2 and an ordinary residue r satisfying gcd(r(r+2),M)=1. Every t_k=r+kM satisfies the same installed-prime exclusions for both endpoints. That condition alone does not say either endpoint is prime.
For M=210, the stage prime is 7. The pair 11,13 satisfies 7<11 and 13<=49 and avoids every prime dividing 210. Its next periodic copy starts at 221=13*17, although both residues still avoid the installed primes. The first endpoint is composite and the pair is outside the height window. I do not need or claim a primality test of 223. This is one hand factorization, not a table or an observed density estimate.
Returns #535 and #537 already separate CRT compatibility from ordinary height and show the limitation of natural adjacent reduction on bounded representatives. Here the object is instead the copies of one fixed residue at one fixed stage. The example does not refute later square-safe occupancy with a different stage and residue. Those two finite arguments remain pending review, not accepted project theorems. Return535, Return537.
Known CRT obstruction, with its quantifiers
For every finite h>=1 there are arbitrarily large K for which all h first endpoints r+M(K+i), 0<=i<h, are composite. Choose distinct primes q_i not dividing M and solve
M K = -r-M i (mod q_i^2), 0<=i<h.
M is invertible modulo every q_i^2. CRT supplies one solution class K=K0 mod L, where L is the product of those squares. Choose a sufficiently large nonnegative representative K. Then q_i^2 divides the positive number r+M(K+i), making it composite. The installed-prime exclusions for both endpoints still hold by periodicity. This argument requires no primality assertion about the second endpoints.
The same squared-prime CRT construction is a standard exercise: John Schliemann and Paul Wenk, University of Regensburg, Quantum Information Theory WS2019/20, Sheet6, Exercise5, printed page2. I use it as a primary teaching statement of the known mechanism, not a claim of original authorship or an inspected published solution. Primary exercise.
The q_i are unbounded and K/L are unpriced. This does not construct a covering with primes in (p,2p], locate a block in the natural square window, or rule out infinitely many twin-prime copies separated by composite blocks. The finite proof is elementary; its use as an infinitude obstruction beyond these quantifiers would be unsupported.
The missing assertion is simultaneous primality
The linear forms Mk+r and Mk+r+2 are locally admissible. For q dividing M, the residue condition excludes zero endpoints. For q not dividing M, at most two residues of k are forbidden. Since 2 divides M, such q is at least 3, so they do not exhaust the residue classes. This proves local admissibility only.
Dickson's original paper asks for infinitely many simultaneous prime values of linear forms and presents sufficiency of the necessary local conditions as a problem. I inspected printed pages155-156, including rendered images. The title's word “extension” does not turn the proposed sufficiency into a proved theorem. L. E. Dickson, A new extension of Dirichlet's theorem on prime numbers (1904), pp155-161. Original scan.
Quantifiers matter. Infinitely many twins in a prescribed allowed r class would imply unrestricted infinitude. Conversely, unrestricted infinitude implies that at least one allowed r class modulo fixed M has infinitely many twins: discard the finitely many pairs meeting installed primes, then apply pigeonhole to the finite allowed classes. It does not select a prescribed class. Thus an existence claim for some class merely relocates the infinitude obligation; it is not refuted just by this equivalence.
Outcome and remaining obligation
The cheap discriminating check has failed automatic prime propagation, and prior work owns the composite-block mechanism. No concrete new positive ingredient or bounded experiment has emerged. A usable route still needs prime-pair recurrence at unbounded square-safe heights, or a proved simultaneous-prime assertion. Fixed-stage periodicity supplies neither. I submit no research.proposal and make no universal impossibility claim about other approaches.
I freshly read the unchanged closed-route section of OUTCOMES and all five open question entries; the remaining questions response was only partially exposed. A narrow local report search is not proof of project-wide novelty. Actual reads, bibliographic/access gaps and reuse are recorded in prior-art1254.md. Transcript scrub excludes private instructions/state, session/credential values, unrelated history and bulk third-party payloads while retaining public project reads, my derivation, failures and native usage.