A different workaround is adding args '-v /var/jenkins-legaci-lab/caches:/var/jenkins-legaci-lab/caches' to the docker{...} declaration in the pipeline.
Like this:

pipeline {
    agent {
        docker {
            label 'docker'
            image 'busybox'
            args '-v /var/jenkins/caches:/var/jenkins/caches'
        }
    }
    stages {
        stage("Test sh script in container") {
            steps {
              sh label: 'Echo "Hello World...', script: 'echo "Hello World!"'
            }
        }
    }
}

• Perhaps this should be solved in some declarative pipeline component?

I confirm it affects me as well:

Jenkins ver. 2.190.1, docker, kubernetes and other plugins: latest stable version.

Jenkins runs on linux (inside a docker container)

Also a problem when running Jenkins not in Docker and executing pipeline on Jenkins master (=> psst!); problem is definitely caused by new "Durable Task" plugin v1.31 bug:

• Latest version of Jenkins core (v2.201) running on Ubuntu 16.04
• The logs seem to indicate the problem already appears when trying to start the Docker container in the pipeline, but maybe the logs are just mangled?
  • See "// !!!" comment in belows build log...
• Build log (proprietary):

  ...
  [Pipeline] // stage
  [Pipeline] stage
  [Pipeline] { (linkchecker)
  [Pipeline] script
  [Pipeline] {
  [Pipeline] withEnv
  [Pipeline] {
  [Pipeline] withDockerRegistry
  [Pipeline] {
  [Pipeline] isUnix
  [Pipeline] sh
  06:37:20  + docker inspect -f . ACME/linkchecker:5
  06:37:20  
  06:37:20  Error: No such object: ACME/linkchecker:5
  06:37:20  06:37:20.408218 durable_task_monitor.go:63: exit status 1              // !!!
  [Pipeline] isUnix
  [Pipeline] sh
  06:37:20  + docker inspect -f . dockerregistry.ACME.com/ACME/linkchecker:5
  06:37:20  .
  [Pipeline] withDockerContainer
  06:37:20  Jenkins does not seem to be running inside a container
  06:37:20  $ docker run -t -d -u 10112:10005 -w /var/lib/jenkins/workspace/Sandbox/ACME.linkCheckerPipeline -v /var/lib/jenkins/workspace/Sandbox/ACME.linkCheckerPipeline:/var/lib/jenkins/workspace/Sandbox/ACME.linkCheckerPipeline:rw,z -v /var/lib/jenkins/workspace/Sandbox/ACME.linkCheckerPipeline@tmp:/var/lib/jenkins/workspace/Sandbox/ACME.linkCheckerPipeline@tmp:rw,z -e ******** -e ******** -e ******** -e ******** -e ******** -e ******** -e ******** -e ******** -e ******** -e ******** -e ******** -e ******** -e ******** -e ******** -e ******** -e ******** -e ******** -e ******** -e ******** -e ******** -e ******** -e ******** -e ******** -e ******** -e ******** -e ******** dockerregistry.ACME.com/ACME/linkchecker:5 cat
  06:37:21  $ docker top 39f784ea27cbf6593fd40c1faaf04948daae94e97eb8ba42517f7c2f5e40c21e -eo pid,comm
  [Pipeline] {
  [Pipeline] sh
  06:42:27  process apparently never started in /var/lib/jenkins/workspace/Sandbox/ACME.linkCheckerPipeline@tmp/durable-aed939a9      // !!!
  06:42:27  (running Jenkins temporarily with -Dorg.jenkinsci.plugins.durabletask.BourneShellScript.LAUNCH_DIAGNOSTICS=true might make the problem clearer)
  [Pipeline] }
  ...
  ERROR: script returned exit code -2
  Finished: FAILURE
  

• Pipeline code (in shared Jenkins pipeline library):

  ...
    void execute(Closure configBody) {
      LinkCheckerRunDSL config = calcConfiguration(configBody)
  
      // Then build, based on the configuration provided:
      script.docker.withRegistry(Constants.ACME_DOCKER_REGISTRY_URL) {
        script.docker.image(config.dockerImage).inside() { c ->
          script.sh 'linkchecker --version'
  ...
  

Having the same issue with the 1.31 version of durable-task plugin. Had to rollback to 1.30.