chore(workflow): trim compute-properties mem comment to drop machine-specific RSS figures

Author: PaulNewling

workflow/src/main.tpl.tengo

@@ -342,9 +342,9 @@ wf.body(func(args) {
 	// stats.json (dataset-level scalars consumed by the info layer — e.g. R11c
 	// median CDR-H3 length per chain).
 	soft := assets.importSoftware("@platforma-open/milaboratories.sequence-properties.software:compute-properties")
-	// Measured optimized footprint of the vectorized Python step: ~4.85 GB peak
-	// RSS per 1M full-paired antibody clones (so ~9.7 GB at 2M). 12GiB covers ~2M
-	// with headroom; cpu(1) stays single-threaded by design.
+	// 12GiB (up from the 4GiB default): the vectorized step materializes full-N
+	// count matrices, so peak memory scales with clone count. cpu(1) is
+	// deliberate — the engine is single-threaded for deterministic output.
 	pyRun := exec.builder().
 		software(soft).
 		mem("12GiB").