hcheck logging to cyclecloud portal is not working correctly

According to nhc logging /var/log/nhc.log, there is an error but hcheck/send_log is not sending it to the cyclecloud portal.

tail /var/log/nhc.log
Running check: "check_hw_physmem 928858388kB 928858388kB 3%"
Running check: "check_hw_swap 0kB 0kB 3%"
Running check: "check_hw_ib 40 mlx5_an0:1"
**ERROR: nhc: Health check failed: check_hw_ib: No IB port mlx5_an0:1 is ACTIVE (LinkUp 40 Gb/sec).**

In NHC DETACHED_MODE, hcheck creates /var/log/healthcheck.json, but it reports no errors (exit-code is 0)?
{
"metadata": {
"azEnvironment": "AzurePublicCloud",
"isHostCompatibilityLayerVm": "false",
"location": "southcentralus",
"name": "xxxxx",
"offer": "ubuntu-hpc",
"osType": "Linux",
"platformFaultDomain": "0",
"platformUpdateDomain": "1",
"provider": "Microsoft.Compute",
"publisher": "microsoft-dsvm",
"resourceGroupName": "slm27u2004hc-xxxx",
"resourceId": "/subscriptions/xxxxx",
"subscriptionId": "xxxx",
"version": "20.04.2022121201",
"vmId": "xxxxx",
"vmScaleSetName": "xxxxx",
"vmSize": "Standard_ND96asr_v4",
"zone": ""
},
"testresults": {
"/opt/cycle/nodehealth/nhc-runner.sh": {
**"exit-code": 0**,
"test-time": 155.6628,
"extra-info": "None",
"message": "Node Health Check detached parent completed successfully (0s).\n"
}
}
}

I think the problem is

hpcadmin@slm27u2004hc-hpc-pg0-1:/opt/cycle/nodehealth$ sudo ./nhc-runner.sh
+ getopts L x
+ OPTIND=0
+ nhc NHC_CHECK_ALL=0 LOGFILE=- MARK_OFFLINE=0
Node Health Check detached parent completed successfully (0s).
**+ exit 0**

Also the hcheck tool does not seem to work when nhc is running in DETACHED_MODE=0 (not detached mode), I get this error message "There was an error in launching the script: Script ./nhc-runner.sh timed out.". ?

Start with /opt/cycle/nodehealth/nhc-runner.sh and the hcheck/send_log path, then compare their behavior with /var/log/nhc.log and /var/log/healthcheck.json in detached and non-detached modes. Reproduce the reported health-check failure and timeout, and trace how the exit code and error are recorded and sent to the CycleCloud portal. Done means failed checks are represented in the JSON and logged to the portal, while successful runs remain distinguishable.

Written by the indexing model from the issue text.