starts getting timeout errors with ECR registeries until pod is restarted

[pdeva]

Describe the bug
seeing timeout errors with ECR from time to time. simply restarting the image-updater pod fixes it. only for the problem to randomly surface hours later...

To Reproduce
this is the relevant config of image updater helm chart:

              - name: config.registries[0].name
                value: "ECR"
              - name: config.registries[0].api_url
                value: "https://xxx.dkr.ecr.us-east-1.amazonaws.com"
              - name: config.registries[0].prefix
                value: "xxx.dkr.ecr.us-east-1.amazonaws.com"
              - name: config.registries[0].default
                value: "true"
              - name: config.registries[0].ping
                value: "true"
              - name: config.registries[0].insecure
                value: "false"
              - name: config.registries[0].credentials
                value: "ext:/scripts/ecr_login.sh"
              - name: config.registries[0].credsexpire
                value: "11h"
              - name: "authScripts.enabled"
                value: "true"
              - name: 'authScripts.scripts.ecr_login\.sh'
                value: >
                  #!/bin/sh

                  aws ecr --region us-east-1 get-authorization-token --output text --query
                  'authorizationData[].authorizationToken' | base64 -d

Expected behavior
if the auth works once, it should work always. shouldnt randomly time out. shouldnt need pod restarts to fix.

Additional context
Add any other context about the problem here.

Version

0.12.2

Logs

Could not set registry endpoint credentials: error executing /scripts/ecr_login.sh: `/scripts/ecr_login.sh` failed timeout after 10s

[pdeva]

seems related to #657