From Irreversibility Memory Kernels to Fine-Grid Stress Testing in Selector-Time Theory: Critical Memory Zones, Residual Signatures, and Robustness in a Toy Model

This article presents the second computational stage of the methodological sequence developed within Selector-Time Theory (STT), continuing the previous work on a minimal irreversibility memory kernel and post-IRREV contextual typification. The objective is not to claim physical validation of Selector-Time Theory, but to transform an ontological hypothesis about irreversibility, residual memory, path dependence, and loss of invertibility into a criticizable, reproducible, adjustable, and progressively refinable computational protocol. The article analyzes Listings 03B, 04, 05, and 06. Listing 03B performs a refined parameter sweep of the memory kernel. Listing 04 builds a toy-model prediction of a critical memory zone. Listing 05 tests the local robustness of that prediction through multiple seeds and controlled perturbations. Listing 06 performs a fine-grid stress test under increasing controlled noise in order to evaluate the stability of the critical zone, the residual peak, the post-critical decay region, and the onset of dynamic locking. The results indicate that the residual signature of the kernel does not grow indefinitely with memory weight. Instead, it organizes into a regime architecture: submemorial region, critical memory band, post-critical decay, and locked memory. Coarse-grid experiments located an initial critical reference near wₘem = 0. 30. The fine-grid stress test refined this result, placing the effective peak of the residual signature approximately in the range 0. 34 <= wₘem <= 0. 35, with an approximate critical band between 0. 22 and 0. 36, and the onset of locking between 0. 37 and 0. 39. The contribution of the work is methodological: it proposes a bridge between an ontological hypothesis about irreversibility and a numerical protocol that can be audited, reproduced, corrected, and eventually compared with known physical models or empirical data. The publication is accompanied by a supplementary ZIP archive containing the Python scripts, bilingual READMEs, CSV, JSON, NPZ, PNG figures, and terminal logs associated with the computational sequence.