[Bug] Wrong number of columns Error

• asyncpg version: 0.23
• PostgreSQL version: 13
• Do you use a PostgreSQL SaaS? If so, which? Can you reproduce
  the issue with a local PostgreSQL install?: Yandex Cloud
• Python version: 3.7
• Platform: Ubuntu 18.04
• Do you use pgbouncer?: yes
• Did you install asyncpg with pip?: no
• If you built asyncpg locally, which version of Cython did you use?: no
• Can the issue be reproduced under both asyncio and
  uvloop?: don't know

We have a web server on Python which is sending queries to the PostgreSQL cluster. After adding a column to the table using a migration without disconnecting I encounter errors with some queries to that table. Example of the query below:

WITH row_info as ( SELECT id FROM UNNEST($1::scheme.affected_table[]) ) DELETE FROM scheme.affected_table AS entities USING row_info WHERE entities.id = row_info.id;

To $1 I pass the list of dictionaries like [{'id': 1, 'field1': 'value1', ...}].

Before the migration or from new connections the query works perfectly. But for queries from web-server I get "wrong number of columns: 11, expected 12". Rebooting the web-server resolves the problem.

I think the problem can be related to Prepared Statements or encoding of Mapping.

The report provides no source file or test entry point. Start by reproducing the UNNEST query after a schema migration with an existing connection, then compare it with a new connection while investigating prepared statements and mapping encoding. Done means the stale-connection query no longer reports a column-count mismatch.