ERROR: bad lexical cast: source type value could not be interpreted as target

[dragonsahead]

Hi, we're from Metabase and a customer of us is getting:

Error executing query: ERROR: bad lexical cast: source type value could not be interpreted as target"

when running the following query

SELECT
  c.column_name AS name,
  c.data_type AS "database-type",
  c.ordinal_position - 1 AS "database-position",
  c.table_schema AS "table-schema",
  c.table_name AS "table-name",
  pk.column_name IS NOT NULL AS "pk?",
  CASE
    WHEN c.remarks <> '' THEN c.remarks
    ELSE NULL
  END AS "field-comment"
FROM
  svv_columns AS c
  LEFT JOIN (
    SELECT
      tc.table_schema,
      tc.table_name,
      kc.column_name
    FROM
      information_schema.table_constraints AS tc
      INNER JOIN information_schema.key_column_usage AS kc
        ON tc.constraint_name = kc.constraint_name
        AND tc.table_schema = kc.table_schema
        AND tc.table_name = kc.table_name
    WHERE
      tc.constraint_type = 'PRIMARY KEY'
  ) AS pk
    ON c.table_schema = pk.table_schema
    AND c.table_name = pk.table_name
    AND c.column_name = pk.column_name
WHERE
  c.table_schema !~ '^information_schema|catalog_history|pg_'
  AND c.table_schema IN (
<schemas here>
  )
ORDER BY
  "table-schema" ASC,
  "table-name" ASC,
  "database-position" ASC;

The schemas parameter is a set of schemas, and there's only one schema that is causing this (every schema is a federated postgres)

The customer also run the query in PyCharm and works, that's why we think this is an issue with the driver