recover original comment

Signed-off-by: Nicolas L. Guidotti <nguidotti@nvidia.com>

Author: nguidotti

cpp/src/mip_heuristics/mip_constants.hpp

@@ -22,8 +22,6 @@
 #define CUOPT_MIP_BATCH_PDLP_REQUIRED_THREAD_COUNT  3
 #define CUOPT_MIP_CLIQUE_CUTS_REQUIRED_THREAD_COUNT 3
 
-// Concurrent LP root solve from inside MIP: barrier runs as a third OMP task alongside PDLP
-// and dual simplex. Disabled below this thread count so the barrier work doesn't overshoot
-// the MIP solver's num_cpu_threads budget (need 1 PDLP + 1 dual simplex + 1 barrier).
-// Stand-alone LP always runs all three concurrently regardless of this gate.
+// MIP-only gate: skip the concurrent barrier when fewer threads are available than this
+// (1 PDLP + 1 dual simplex + 1 barrier). Stand-alone LP always runs all three.
 #define CUOPT_CONCURRENT_LP_BARRIER_REQUIRED_THREAD_COUNT 3