CI - Extracted cache is empty if jobs aren't started in same time

Hello everyone, I have a problem with the cache system in my CI process. I’m compiling a program for two plateforms, one a Rapsberry PI and the other for x86.

All jobs are running on the same PC with gitlab-runner.

When the jobs are started at the same time and are at the same stage, there isn’t any problem. The cache is extracted and not empty.
But when they aren’t started at the same time, because one runner thread is already busy, the cache extrated is empty in the second job run.

The extracted information in the fail job is:

Running with gitlab-runner 10.5.0 (10.5.0)
  on ci-pc021 13b403e0
Using Docker executor with image rdh.mag.com:5000/app-ubuntu-16.04 ...
Pulling docker image rdh.mag.com:5000/app-ubuntu-16.04 ...
Using docker image sha256:230c94db1fe54daef116d88aabb1e72ee2fe6a4083546a26630d8e1877b4d21a for rdh.mag.com:5000/app-ubuntu-16.04 ...
Running on runner-13b403e0-project-929-concurrent-0 via pc021...
Fetching changes...
Removing project/workingdir/
HEAD is now at 30fc31b Merge branch 'release/0.3.0'
Checking out 30fc31b9 as master...
Skipping Git submodules setup
Checking cache for RasPi2-master-3...
Successfully extracted cache

The extracted information in the previous successful job run and without error is:

Running with gitlab-runner 10.5.0 (10.5.0)
  on ci-pc021 13b403e0
Using Docker executor with image rdh.magellium.com:5000/msaf-ubuntu-16.04 ...
Pulling docker image rdh.magellium.com:5000/msaf-ubuntu-16.04 ...
Using docker image sha256:230c94db1fe54daef116d88aabb1e72ee2fe6a4083546a26630d8e1877b4d21a for rdh.magellium.com:5000/msaf-ubuntu-16.04 ...
section_start:1568647319:prepare_script
Running on runner-13b403e0-project-929-concurrent-0 via pc021...
Fetching changes...
Removing DataConsoleWriter-COTS-ubuntu-16.04.tar.gz
Removing cots/
Removing cots_sources/
HEAD is now at 30fc31b Merge branch 'release/0.3.0'
Checking out 30fc31b9 as master...
Skipping Git submodules setup
Checking cache for x86-master-3...
Successfully extracted cache

The configuration of the runner is (config.toml):

concurrent = 2
check_interval = 3

[[runners]]
  name = "ci-pc021"
  url = "https://gitana.mag.com"
  token = "13b403e0e81ed09358b5d4c4d5403d"
  executor = "docker"
  [runners.docker]
    tls_verify = false
    image = "ubuntu:16.04"
    dns = ["172.0.1.2", "172.0.1.3"]
    privileged = true
    disable_cache = false
    volumes = ["/cache"]
    shm_size = 0
  [runners.cache]

I don’t understand why. Could it be because the runner runner-13b403e0-project-929-concurrent-0 is being used for the two jobs at the same time ?
Also, I can’t use artefact because the data exceeds the limit to be on it.

Thank you for your help.

Fixed with the last GitLab update.