Context, I have spent some time to dive into the product backlog labeling health:

1. I create filters that allows us to find all the issues without a product scope label: https://www.notion.so/mui-org/product-GitHub-community-issues-PRs-Tier-1-12a84fdf50e44595afc55343dac00fca?source=copy_link#24bcbfe7b66080f19ad0d207c9e5ce1a. Then fixed them to get to zero.
   But I gave up on covering material-ui repo, it would have taken a whole day 😀.
2. I fixed all the issues with more than one label type. From which, it seems clear that expected behavior and duplicate label should be "type" labels because they are mutually exclusive with the other type labels. And with this larger diversity of types labels, adding the type: prefix now feels like a clear win.

However, it feels like there are too many of them that have the wrong type, e.g.:

• [data grid] custom pagination #18431
• [data grid][8.9.0] Attempted import error: 'useRunOncePerLoop' is not exported from '@mui/x-data-grid-pro' (imported as 'useRunOncePerLoop'). #18847
• [data grid] Row-spanning works differently with getRowHeight() #18404

So I think we should normalize back to what the other repositories do. We tried with mui/mui-x, but it looks like the probabilities are something like this:

• user developers apply the right label: 40%
• maintainer see that they needs to apply the right label when there is one applied by default 50%
• maintainer see that they needs to apply the right label when there is none applied by default 95%
• maintainer apply the right label: 95%

Then we can apply the math with ⬆️ values:

1. Label by default, probability to get the right label = 0.40 * 0.50 + 0.40 * 0.50 * 0.95 + 0.60 * 0.50 * 0.95 = 0.68
2. No labels by default, probability to get the right label = 0.95 * 0.95 = 0.90 🏆