Among the oldest forms of written heritage in Asia, palm leaf manuscripts preserve centuries of cultural, religious, literary, and scientific knowledge. However, their analysis using modern document imaging techniques remains limited due to challenges such as script diversity, complex layouts, and physical degradation. While regional research efforts are growing, publicly available annotated datasets remain scarce and fragmented. Notably, many of these ancient, low-resource scripts exhibit unique linguistic characteristics while sharing common structural layout features. In this paper, we introduce PALM-LAY, the first unified, cross-regional annotated dataset specifically designed for the layout analysis of historical palm leaf manuscripts. The dataset includes over 566 pages and more than 6,000 annotated regions, covering six distinct scripts: Tamil, Kambaramayanam, Jathakam, Khmer, Sundanese, and Balinese. These samples were carefully curated from multiple publicly available collections and annotated using a unified layout schema. Each image is annotated with seven region categories: MainRegion, TextLineRegion, ParagraphRegion, SymbolicMark, PhysicalDamage, Illustration, and Other, capturing both traditional manuscript structures and modern archival additions. We benchmark several state-of-the-art object detection models to evaluate layout analysis performance across scripts. The results highlight the challenges of cross-script generalization and underscore the need for culturally informed, layout-aware models. PALM-LAY is intended to advance research in multilingual document understanding and support the digital preservation of endangered manuscript traditions. The dataset and documentation will be publicly released at: https://github.com/back-kh/PALM-LAY .