Apologies if this has already been referenced elsewhere, or is already the case.
This might be prone to abuse. A malicious DT1 member right now can't unilaterally kick someone out from DT2, they need one or more other DT1 members to exclude a DT2 member. In your proposed scenario a DT1 member could keep adding sockpuppets (or just random users who happen to exclude the targeted DT2 member) until they reach their goal.
Fair point, though I think the issue you've described here would be the problem with untrustworthy DT1 members rather than anything else. For this reason and your argument, I believe this would be good to additionally "root out" any untrustworthy DT1 members who are abusing their power at the same time. You do have a point though, I'm not denying that.
A solution to this could be an addition to this algorithm that DT2 members (who are as a "majority" excluding a DT2 member) must be trusted from
different DT1 members. This would then require a conspiracy within DT1 to exclude a DT2 member, which would be less of a conspiracy and more of a co-ordinated effort.