The rationale etc. didn't change in v2, so see
https://lore.kernel.org/r/20230510160428.175409-1-johannes@sipsolutions.net

In v2 I dropped the workqueue locking patch and consequently
made new wiphy_work and wiphy_delayed_work things.

I kept the pause/resume for the workqueue because not that I
had it, it seemed simpler than keeping track of "should it be
paused" in a separate bool variable to not queue the work when
it shouldn't be ...

Still just serves to illustrate things, I didn't really test it.

johannes