All of them? Even the ones that cannot be expressed by administrative boundaries but yet are verifiable on the ground?
My understanding is that OSM doesn’t strictly need a giant boundary relation for every time zone, because Timezone Boundary Builder can generate them from a combination of time zone boundaries and administrative boundaries.
In any case, now that OpenHistoricalMap has imported a comprehensive dataset of historical U.S. county boundaries, I guess we can look into mapping the U.S. timezone boundaries there using the county boundaries as a foundation, and also investigate doing something similar for the rest of the world, so that tz database users will continue to have access to useful boundary data. This could be a positive outcome, since the tz database also keeps track of historical changes to time zones that would be out of scope for OSM but well within scope for OHM.