[system test] memcached crashed when add node back in online upgrade from 1.8.1 to 2.0.0

Install couchbase server 1.8.1 on 4 nodes cluster with default installation path and custom data path.
10.3.2.10
10.3.2.11
10.3.2.12
10.3.2.16
10.3.2.75

Create 3 buckets, one default, one sasl and one dedicated port bucket.
Load 20K items to each bucket.
Maintain load about 3K at each bucket.

Install couchbase server 2.0.0-1974 on node 10.3.2.76 with default installation path and custom data path.
Swap rebalance (add node 76 and remove node 75). Rebalance done.

Install couchbase server 2.0.0-1974 on node 10.3.2.77 with default installation path and custom data path.
Swap rebalance (add node 77 and remove node 16). Rebalance done.

Uninstall couchbase server 1.8.1 on node 16 and 75
Install couchbase server 2.0.0-1974 on node 16 and 75 with custom installation path (c:/one) and custom data path (c:/data)
Swap rebalance (add node 16 and 75 to cluster, remove node 11 and 12)
Rebalance failed in few minutes

I see a lot of error in log page with error

Port server memcached on node 'ns_1@10.3.2.16' exited with status 1. Restarting. Messages: Wed Dec 05 16:16:29.949804 Pacific Standard Time 3: Error initializing sasl.

Filed bug MB-7369

Then clean install node 16 and node 75 with default installation path (c:/Program Files/Couchbase/Server/)
Add node 16 and 75 back to cluster
10.3.2.10
10.3.2.11
10.3.2.12
10.3.2.76
10.3.2.77

Remove node 11 and 12 to do swap rebalance.
Rebalance failed. Check memcached on node 16 and 75, I see memcached crashed continuously
I see the bucket with dedicated port (port 11207) is in pending state as in capture

The cluster is in failed state now.