Backporting regression fix from master to 9.x about modules

[benjamingr]

Are there objections to landing this? As it had been landed on master prior to us deciding that we wanted to discuss a code freeze, I think it makes sense to backport.

In general, I think that:

• The consensus is against a code freeze at the moment, while we agreed that the current implementation might break in incompatible ways.
• We should default to the current status, as we're not a chartered working group - even if we wanted to object to landing something we should likely do so in the "regular" TSC approved way (in the PR)

[jdalton]

I don't see this as controversial.
That said, bugs in something experimental are nothing new so I don't see backporting as a must either.

[MylesBorins]

Thanks for chiming in folks. I landed the backport. We can discuss further in the meeting next week