What is the issue with the HTML Standard?
First discussed in issue #11845 (this in particular.
If there is a requirement that onsuccess should fire after reactions to the committed promise this should be made explicit in spec. I think onerror doesn't have this requirement, so possibly Gecko's solution isn't entirely optional and only the success callback should be a reaction to the committed promise, if such exists. Implicitly betting that the extra lap on the microtask queue fixes this is not a solution, especially for fast failure handling and non-method tracker navigations.
In the end these order issues are an inherent problem in the difference in nature between events and promises.