diff --git a/README.md b/README.md
index 806e037d1..e872da930 100644
--- a/README.md
+++ b/README.md
@@ -223,8 +223,20 @@ You can load this into your local Docker client by running:
 `bazel run my/image:helloworld`.
 
 For the `lang_image` targets, this will also **run** the
-container to maximize compatibility with `lang_binary` rules.  You can suppress
-this behavior by passing the single flag: `bazel run :foo -- --norun`
+container using `docker run` to maximize compatibility with `lang_binary` rules.
+
+Arguments to this command are forwarded to docker, meaning the command
+
+```bash
+bazel run my/image:helloworld -- -p 8080:80 -- arg0
+```
+
+performs the following steps:
+* load the `my/image:helloworld` target into your local Docker client
+* start a container using this image where `arg0` is passed to the image entrypoint
+* port forward 8080 on the host to port 80 on the container, as per `docker run` documentation
+
+You can suppress this behavior by passing the single flag: `bazel run :foo -- --norun`
 
 Alternatively, you can build a `docker load` compatible bundle with:
 `bazel build my/image:helloworld.tar`.  This will produce the file:
diff --git a/container/incremental_load.sh.tpl b/container/incremental_load.sh.tpl
index 438861b2a..3b616d289 100644
--- a/container/incremental_load.sh.tpl
+++ b/container/incremental_load.sh.tpl
@@ -242,11 +242,41 @@ function read_variables() {
 # An optional "docker run" statement for invoking a loaded container.
 # This is not executed if the single argument --norun is passed or
 # no run_statements are generated (in which case, 'run' is 'False').
-if [[ "a$*" != "a--norun" && "%{run}" == "True" ]]; then
+if [[ "%{run}" == "True" ]]; then
+  docker_args=()
+  container_args=()
+
+  # Search remaining params looking for docker and container args.
+  #
+  # It is assumed that they will follow the pattern:
+  # [dockerargs...] -- [container args...]
+  #
+  # "--norun" is treated as a "virtual" additional parameter to
+  # "docker run", since it cannot conflict with any "docker run"
+  # arguments.  If "--norun" needs to be passed to the container,
+  # it can be safely placed after "--".
+  while test $# -gt 0
+  do
+      case "$1" in
+          --norun) # norun as a "docker run" option means exit
+              exit
+              ;;
+          --) # divider between docker and container args
+              shift
+              container_args=("$@")
+              break
+              ;;
+          *)  # potential "docker run" option
+              docker_args+=("$1")
+              shift
+              ;;
+      esac
+  done
+
   # Once we've loaded the images for all layers, we no longer need the temporary files on disk.
   # We can clean up before we exec docker, since the exit handler will no longer run.
   cleanup
 
   # This generated and injected by docker_*.
-  exec %{run_statements}
+  exec %{run_statement} "${docker_args[@]}" "%{run_tag}" "${container_args[@]}"
 fi
diff --git a/container/layer_tools.bzl b/container/layer_tools.bzl
index 1670d82ce..dcd093e87 100644
--- a/container/layer_tools.bzl
+++ b/container/layer_tools.bzl
@@ -220,18 +220,21 @@ def incremental_load(
 
     toolchain_info = ctx.toolchains["@io_bazel_rules_docker//toolchains/docker:toolchain_type"].info
 
-    # Default to interactively launching the container,
-    # and cleaning up when it exits.
-
-    run_flags = run_flags or "-i --rm"
-
     if len(images) > 1 and run:
         fail("Bazel run does not currently support execution of " +
              "multiple containers (only loading).")
 
+    # Default to interactively launching the container, and cleaning up when
+    # it exits. These template variables are unused if "run" is not set, so
+    # it is harmless to always define them as a function of the first image.
+    run_flags = run_flags or "-i --rm"
+    run_statement = "\"${DOCKER}\" ${DOCKER_FLAGS} run %s" % run_flags
+    run_tag = images.keys()[0]
+    if stamp:
+        run_tag = run_tag.replace("{", "${")
+
     load_statements = []
     tag_statements = []
-    run_statements = []
 
     # TODO(mattmoor): Consider adding cleanup_statements.
     for tag in images:
@@ -271,11 +274,6 @@ def incremental_load(
                 _get_runfile_path(ctx, image["config_digest"]),
             ),
         )
-        if run:
-            # Args are embedded into the image, so omitted here.
-            run_statements.append(
-                "\"${DOCKER}\" ${DOCKER_FLAGS} run %s %s \"$@\"" % (run_flags, tag_reference),
-            )
 
     ctx.actions.expand_template(
         template = ctx.file.incremental_load_template,
@@ -283,7 +281,7 @@ def incremental_load(
             "%{docker_flags}": " ".join(toolchain_info.docker_flags),
             "%{docker_tool_path}": docker_path(toolchain_info),
             "%{load_statements}": "\n".join(load_statements),
-            "%{run_statements}": "\n".join(run_statements),
+            "%{run_statement}": run_statement,
             "%{run}": str(run),
             # If this rule involves stamp variables than load them as bash
             # variables, and turn references to them into bash variable