Good question! By and large, some of the following would count as a "strike":
Poor moderation
Excessive spam
Most importantly: Making non-backwards compatible changes to the ActivityPub specification that may lead to the classic "Embrace, extend, extinguish" situation
Threads as it exists now is poorly moderated in that it is 90% shameless self-promotional commercial content.
If your instance doesn't have a policy against horrible undisclosed commercial promotion, I want nothing to do with it. If it does, Threads already has hundreds of thousands of strikes.
I’d say any automated/integrated effort to direct users of federated instances to the threads site to view content should count as a strike. (Such as needing to go directly to the threads site to view an image that could be easily posted anyway.)
So should any automated/integrated effort to encourage users to make their own threads account. (Such as needing an account to visit this link or view this image.)
Any attempt to coerce non threads users to sign any sort of agreement or TOS with threads.
As well as any data collection on non threads users. Merely interacting with a federated threads account should not entitle meta to any data collection of that user.
is it even possible to federate with them without receiving thousands of posts per minute of barely moderated content that would drown everything else in the feed?
Most importantly: Making non-backwards compatible changes
That's the shift into "Extend" - they won't do this until their Embrace phase has enmeshed their users with other fediverse users so that defederation affects people's subscriptions.
I'm actually hoping Google or Microsoft or Apple etc create a compatible activity pub based service. That would create balance and make 'extend' problematic for meta.
I also think we need state actors and universities to start using Mastodon (not Twitter) - again that would make it difficult for Meta to deviate from standards.