Top of my table: lasers/EML #1 (very high conviction — "most direct optical bottleneck" for 800G/1.6T and co-packaged ramps), signal processors #2 (power-performance gate), InP substrate & epitaxy #3-4, precision packaging #5 (manufacturing-yield gate). Thirteen ranks across six categories, each a dated April 2026 hypothesis flagged to-be-validated until primary evidence promotes it.