Unknown stop on Ubuntu 16/18 images

We need to test with old compilers in Boost projects so we are using e.g Ubuntu 18 via a docker image.

As Node 20/24 builds uses a glibc newer than 2.17 (used on those images) we use a workaround employed after https://github.com/actions/checkout/issues/1590 updated for Node 24. Basically:

jobs:
  posix:
    runs-on: ${{matrix.os}}
    container:
      image: ${{matrix.container}}
      volumes:
        - /node20_217:/node20_217:rw,rshared
        - /node24_217:/node24_217:rw,rshared
        - ${{ startsWith(matrix.container, 'ubuntu:1') && '/node20_217:/__e/node20:ro,rshared' || ' ' }}
        - ${{ startsWith(matrix.container, 'ubuntu:1') && '/node24_217:/__e/node24:ro,rshared' || ' ' }}

And a step that extracts unofficial node 20/24 builds into /node20217

This works for actions/checkout but with actions/cache the action just stops with an error but without any output on why. The last message is:

Cache hit for restore-key: ubuntu-latest-ubuntu:16.04-gcc-4.8-9267b441bfa8cb4a388da922525d6910f458fae0

See https://github.com/boostorg/boost-ci/actions/runs/28161775001/job/83404649058?pr=329#step:5:24

Any ideas what can cause this?
Is there any way to get more/debug output of this action?

Start with the posix job's container setup and the actions/cache step in the linked Boost CI run; compare its behavior with actions/checkout under the same Node 20/24 and glibc workaround. Check the cache action's final restore-key message and available workflow debug output, then reproduce or narrow the silent stop on Ubuntu 16/18. Done means the cause is identified and the action reports or handles it with a verified workflow result.