connection.close is throwing protocol error

asyncpg version: 0.18.1
PostgreSQL version: 9.5.10
Do you use a PostgreSQL SaaS? If so, which? Can you reproduce
the issue with a local PostgreSQL install?: using google cloud sql pg instance
Python version: 3.6.5
Platform: Ubuntu
Do you use pgbouncer?: No
Did you install asyncpg with pip?: Yes
If you built asyncpg locally, which version of Cython did you use?:
Can the issue be reproduced under both asyncio and
uvloop?: No

I have been facing this error for quite sometime intermittently `asyncpg.exceptions._base.InternalClientError: got result for unknown protocol state 0`

The entire stack trace:
```
File "/repo/server/schema_db.py", line 345, in read
await connection.close()

File "/usr/local/lib/python3.6/dist-packages/asyncpg/connection.py", line 1082, in close
await self._protocol.close(timeout)

File "asyncpg/protocol/protocol.pyx", line 560, in close

File "asyncpg/protocol/protocol.pyx", line 814, in asyncpg.protocol.protocol.BaseProtocol._dispatch_result

asyncpg.exceptions._base.InternalClientError: got result for unknown protocol state 0
```

Can someone explain when this exception is thrown and for what reason?
If i know the reason this exception is thrown, i can figure out the issue myself.
I'm not able to reproduce locally or in my stage environment.

Start with the failing await connection.close() at schema_db.py:345, then inspect asyncpg/protocol/protocol.pyx at close and BaseProtocol._dispatch_result, using the reported stack trace and version details. Determine the intermittent conditions that produce the unknown protocol state with the Google Cloud SQL PostgreSQL instance; the issue is done when the cause and a reliable reproduction or verification path are established.

Written by the indexing model from the issue text.