A known issue in an upstream component results in an error and crash when too many layers are present on our cloud builders and we attempt to list the metadata for garbage collection.
We have encountered this before, and in the past we lowered the threshold for garbage collection to avoid significant layer buildup.
However, this does not address the root cause and it is still possible to exceed the metadata limit with many smaller layers in the local engine cache.
To avoid a repeat of this incident, will revisit the upstream issue and open a pull request with a fix ourselves.