Centromeres are essential for chromosome segregation, yet their organisation and evolution remain poorly understood in holocentric species, where kinetochore activity is distributed along entire chromosomes. While monocentric centromeres are often structured by megabase-sized satellite arrays, the role of repetitive DNA in holocentric systems remains enigmatic. Here, we analyse the dynamics of centromeric Tyba satellite DNA repeats and transposable elements across a chromosome-scale pangenome comprising 56 long-read haplotype assemblies from 20 Rhynchospora species, a plant genus with repeat-based holocentromeres. We identify over 4.6 million monomers of the Tyba satellite repeat, arranged into 43,400 discrete arrays that span all chromosomes. CENH3 ChIP-seq reveals that, unexpectedly, the same Tyba satellite defines holocentromere across the entire genus, demonstrating deep conservation of centromeric DNA over over 40 million years despite extensive karyotype evolution and centromere array turnover. We show that Tyba arrays function as modular centromeric units whose number and spacing, but not size, scale with chromosome length. Tyba sequence diversity recapitulates species phylogeny, while higher-order repeat formation and antagonism with transposable elements shape array turnover. A novel synteny-aware algorithm reveals rapid gain, loss, and rearrangement of arrays across homologous chromosomes. Using cytogenetics and polymer simulations, we demonstrate that inter-array spacing governs chromatin loop length and chromatid thickness, linking repeat-based holocentromere organisation directly to chromosome mechanics. Our findings uncover a scalable, modular logic for holocentromere function and establish a framework for understanding the plasticity of repeat-based centromere evolution and genome architecture in eukaryotes.
