Example here: https://github.com/gotson/sqlite-jdbc/blob/0396dd33095d2045adef229a4530240d2f1f8a4f/.github/workflows/test.yml#L76

I'm having the same problem. It does not fail but the cache is invalidated and rebuilt because of different node versions coming from the matrix.
Example: https://github.com/giggio/node-chromedriver/runs/5439981596?check_suite_focus=true

I experience the same problem: https://github.com/apache/avro/blob/a6c82ac6f4d316e769059adab39a9b39b5ea9f8c/.github/workflows/test-arm64.yml

My workflow has several jobs which run the action in their steps. Since all jobs share the same Docker image they invalidate the image for each other.

I have worked around this issue by overwriting GITHUB_WORKFLOW env var: apache/avro@c635b9a

I have worked around this issue by overwriting GITHUB_WORKFLOW env var: apache/avro@c635b9a

@martin-g this worked! I am using a matrix, so I had to use matrix values when setting GITHUB_WORKFLOW. Thanks for the tip!
giggio/node-chromedriver@8255bb5

On the contrary I would like to use the same cache for different workflows. Instead of overwriting GITHUB_WORKFLOW I would want to allow specifying the full container name:

diff --git a/src/run-on-arch.js b/src/run-on-arch.js
index aeca1b2..ab4362c 100644
--- a/src/run-on-arch.js
+++ b/src/run-on-arch.js
@@ -100,7 +100,7 @@ async function main() {
   }
 
   // Generate a container name slug unique to this workflow
-  const containerName = slug([
+  const containerName = slug(core.getInput('containerName') || [
     'run-on-arch', env.GITHUB_REPOSITORY, env.GITHUB_WORKFLOW,
     arch, distro,
   ].join('-'));

That way more complex usecases could always specify the exact name they want to use and less complex ones can use the default. This solution should go well with #100 and allow a lot more flexibility.