madgraph crashes without an error code

I am generating tth events in madgraph with a slightly modified version of the higgs characterisation UFO model. Specifically I need to generate a large number (~1000) of small batches with different model parameters, which I do by generating a madgraph proc card that continously updates two model parameters and re-launches the generation (I can't use scan, since I need to update two parameters together and don't want to do a grid search over these). When I run madgraph however, it crashes at exactly run 811 (this seems to consistent over order 100 identical runs launched at the same time), partially with a proper return code, such that my configured pipeline stops here and partially without a proper return code, leaving the pipeline unable to detect a failure. The one without a proper return code has the following error:

```
"launch" with error:
RuntimeError : can't start new thread
```

and with a proper return code I get

```
Command "generate_events run_811" interrupted with error:
RuntimeError : can't start new thread
```

Could this be an "out of resource" error and how can I circumvent that? How can I force madgraph to always return a non-zero return code if it fails?

Note: This is the template for my proc card:

```
#************************************************************
#* MadGraph5_aMC@NLO *
#* *
#* * * *
#* * * * * *
#* * * * * 5 * * * * *
#* * * * * *
#* * * *
#* *
#* *
#* VERSION 3.2.0 2021-08-22 *
#* *
#* The MadGraph5_aMC@NLO Development Team - Find us at *
#* https://server06.fynu.ucl.ac.be/projects/madgraph *
#* *
#************************************************************
#* *
#* Command File for MadGraph5_aMC@NLO *
#* *
#* run as ./bin/mg5_aMC filename *
#* *
#************************************************************

# define model
import model {{ model_path }}

# define process
define l+ = e+ mu+
define nu = vl vl~
# ensure that the tops, ws and the higgs are on-shell as defined by bwcutoff;
# QNP=0 removes new physics couplings from the corresponding decays
generate p p > t t~ x0, x0 > a a, (t > w+ b QNP=0, w+ > l+ nu QNP=0), (t~ > w- b~ QNP=0, w- > j j QNP=0)
output {{ output_dir }}
launch

# simulation after hard process
shower=Pythia8
detector=off
done

# set run parameters
set run_card run_tag main
set run_card nevents {{ event_count }}
set run_card iseed {{ seed }}
set run_card lpp1 1 # proton beam
set run_card lpp2 1 # proton beam
set run_card ebeam1 6500
set run_card ebeam2 6500

# set pythia parameters
set pythia8_card PartonLevel:MPI off
set pythia8_card PartonLevel:ISR on
set pythia8_card PartonLevel:FSR off
set pythia8_card HadronLevel:all on

# set model parameters
#set param_card cosa scan: [{{ ",".join(cos_alpha_prior) }}]
#set param_card kHtt scan: [{{ ",".join(kappa_higgs_top_prior) }}]
#set param_card kAtt scan: [{{ ",".join(kappa_pseudo_higgs_top_prior) }}]

set param_card cosa {{ cosa_prior[0] }}
set param_card kHtt {{ kappa_higgs_top_prior[0] }}
set param_card kAtt {{ kappa_pseudo_higgs_top_prior[0] }}
done

{% for i in range(kappa_higgs_top_prior[1:]|length) %}
launch
done

set param_card cosa {{ cosa_prior[1:][i] }}
set param_card kHtt {{ kappa_higgs_top_prior[1:][i] }}
set param_card kAtt {{ kappa_pseudo_higgs_top_prior[1:][i] }}
done
{% endfor %}
```
I would also post the full log files, however I'm not sure how to attach files here and they are pretty long. I am using madgraph version 3.1.0.