Regression in RA Msg4

We encounter multiple RA attempts that fail at Msg4 in the RAN-SA-B200-Module-SABOX-Container CI pipeline (as a consequence of this regression, pipeline often fails on the re-transmission check). After multiple attempts, UE usually succeeds with RA and attach to the gNB. This issue appears in 40 MHz TDD SA with SC-FDMA and 40 MHz TDD F1+E1 SA test scenarios. You can see this behavior in the following run:
https://jenkins-oai.eurecom.fr/job/RAN-SA-B200-Module-SABOX-Container/6236/

Upon investigation, I suspect that this regression may be related to MR !2876 (merged).

I tested a modification to reduce the number of max HARQ processes: fd7bbfe9.

diff --git a/common/utils/nr/nr_common.h b/common/utils/nr/nr_common.h
index 94102a858f..fcdcab0a62 100644
--- a/common/utils/nr/nr_common.h
+++ b/common/utils/nr/nr_common.h
@@ -42,7 +42,7 @@
 #define MAX_NUM_BEAM_PERIODS 4
 #define MAX_BWP_SIZE 275
 #define NR_MAX_NUM_BWP 4
-#define NR_MAX_HARQ_PROCESSES 32
+#define NR_MAX_HARQ_PROCESSES 16
 #define NR_NB_REG_PER_CCE 6
 #define NR_NB_SC_PER_RB 12

With this adjustment, the issue with multiple RA attempts no longer occurs, as seen in run: https://jenkins-oai.eurecom.fr/job/RAN-SA-B200-Module-SABOX-Container/6256/

@Raghavendra, could you please review this issue?

@schmidtr, if you have additional insights or observations related to this behavior, please share them.