Memory optimization by accessing exit codes of retries beyond lastRetry

[saikonen]

I'm trying to implement some conditional logic into my workflow template that will increase the memory request for the pods based on past failures of a task.
The simple case where I increase the memory on every failure is straightforward enough

metadata:
  name: retry-podspecpatch-test
  namespace: default
spec:
  templates:
    - name: OOMTest
      inputs: {}
      outputs: {}
      metadata: {}
      dag:
        tasks:
          - name: start
            template: retrystep
        failFast: true
    - name: retrystep
      inputs: {}
      outputs: {}
      metadata: {}
      container:
        name: retrycont
        image: python:alpine3.23
        command:
          - python
          - -c
        args:
          - import random; import sys; exit_code = random.choice([1, 137]);
            sys.exit(exit_code)
        resources:
          requests:
            cpu: "1"
            memory: 512M
      retryStrategy:
        limit: "5"
        retryPolicy: Always
      podSpecPatch: '{"containers": [{"name": "main", "resources": {"requests":
        {"memory": "{{= [1024, 2048, 4096][min(asInt(retries),2)]}}"}}}]}'
  entrypoint: OOMTest
  arguments: {}

What I am aiming for is a more refined condition, where the value only gets incremented if retries exit with 137. Scouring the docs it does not seem to be possible to access the exit codes of all past retries of a task, and lastRetry.exitCode is not enough info to determine which value to pick from the array.

• Is there a way to access all previous exit codes of task retries, so I could construct an index to pick the correct value with?
• Failing to do so, is there a way to build an accumulator between retries, that can be referenced in the podSpecPatch section of the template? From what I can tell, self-referencing steps seem to only lead to errors in the template