XAttr query fails after upgrade

1. Spin up a two node 4.6.5 cluster
2. Load 3000 docs
3. Offline upgrade the cluster to 5.5.0-2907
4. Wait for buckets to be ready
5. Write system xattr data to every doc using python sdk
6. Execute the following rest calls to create proper indexes and query the xattrs:

   curl -v http://172.23.104.105:8093/query/service -d 'statement=CREATE INDEX idx1 ON default(meta().id, meta().xattrs._system1, join_day) USING GSI'

   curl -v http://172.23.104.105:8093/query/service -d 'statement=EXPLAIN SELECT meta().xattrs._system1 FROM default where meta().id is not missing'

   curl -v http://172.23.104.105:8093/query/service -d 'statement=SELECT meta().xattrs._system1 FROM default where meta().id is not missing'

   curl -v http://172.23.104.105:8093/query/service -d 'statement=SELECT meta().xattrs._system1 FROM default use index(`#primary`) where meta().id is not missing'

The last query here with return an error: 

{

"requestID": "cfc4741b-fd05-4e21-8224-3ff45893924c",

"signature": \{"_system1":"json"},

"results": [

],

"errors": [\{"code":12008,"msg":"Error performing bulk get operation  - cause: \{1 errors, starting with MCResponse status=NOT_SUPPORTED, opcode=SUBDOC_MULTI_LOOKUP, opaque=1, msg: }"}],

"status": "errors",

"metrics": \{"elapsedTime": "7.681754ms","executionTime": "7.55602ms","resultCount": 0,"resultSize": 0,"errorCount": 1}