We're doing ACM perf/scale test to deploy 3672 SNOs with ACM 2.11.0.
When deploying OCP 4.15.z, we can got 100% SNOs deployed successfully. But when deploying 4.16.0-rc build, we constantly get about 0.5% deployment failure, which is around 15 of 3672 SNOs. The failure are the same for all 15 SNOs: api server is not reachable after the installation.
one of the SNO's install logs are uploaded to https://drive.google.com/drive/folders/1wKsymDf8-8rvzSPipuXBugQBK3RvUYVA?ths=true
You can also find the output of "journalctl -b -f -u kubelet.service" and "journalctl -b -f -u crio.service" there shown as kubelet.service.log and crio.service.log

the  above logs are from 4.16.0-rc4

1.
2.
3.