Looks like this is related to / fixes:

• Swarm mode cluster fails and does not recover after first node failure #26038
• Docker fails to start: panic: runtime error: index out of range #24315 (which is fixed by Vendoring boltdb @fff57c100 #26607 on master)
• Docker Swarm - service resize during an update can leave inconsistent versions  #26617

@aaronlehmann can you check if those are resolved by this? Also (as I see they're still open); do we need a separate PR to fix this in master?

@cpuguy83 that's my fault 😞 #26707 (comment). And, apparently the DCO check does a diff compared to "master", and checks every commit since (not just the commits in the PR)

Should we rewrite history on the 1.12.x branch? Or add a commit that disables this check?

#26038

Let's keep this open until we confirm that the fixes here help.

#24315

Already closed (is that correct? milestone is 1.12.2 and the docker PR hasn't been backported yet)

#26617

Will close after this is merged.

Also (as I see they're still open); do we need a separate PR to fix this in master?

No, #26538 just vendored swarmkit.

Should we rewrite history on the 1.12.x branch? Or add a commit that disables this check?

Just discussed with @icecrime and that's probably the best option; I'll have a look later tonight, and will let you know when to rebase/update your PR.

@aaronlehmann should be fixed now #26707 (comment)

Rebased.

"lgtm", but ping @LK4D4 to double check if all required fixes from SwarmKit are here

I've updated this to include two more swarmkit PRs:

• fix panic in portallocator and deallocate endpoints when update with none EndpointSpec swarmkit#1481: fix panic in portallocator and deallocate endpoints when update with none EndpointSpec
• validate duplicate published port in ServiceSpec swarmkit#1495: validate duplicate published port in ServiceSpec

LGTM