When using HBMs with:
<id name="accountId" column="`accountId`" type="long">
<generator class="identity" />
</id>

hibernate.dialect=org.hibernate.dialect.PostgreSQLDialect

1. PostgreSQL (speed startup), not sure of relevance to the bug report
   hibernate.temp.use_jdbc_metadata_defaults=false
2. as of 4.2.4 this only does tableName's
   hibernate.globally_quoted_identifiers=true

Hibernate produces SQL like: select currval('"master"."foo_Account"_"accountId"_seq')

The correct format would be:
select currval('"master"."foo_Account_accountId_seq"')

That is the concept of an identifier being the sequence name is a single construct that should be quoted. The use of apostrophe around the outer most part of the value is correct. The expression inside the string need to be a valid Postgres identifier expression.

I would guess hibernate is building the default sequence name out of the tableName + "_" + columnName + "_seq" format that postgres uses for SERIAL/BIGSERIAL.

What should happen is that if either the tableName or the columnName is using quoting, then the entire string should be quoted as a single identifer and the unquoted tableName and columnName be used inside.

StringBuilder sb = new StringBuilder();
sb.append(tableName.getUnquotedName()).append("_");
sb.append(columnName.getUnquotedName()).append("_seq");
if(tableName.isQuoted() || columnName.isQuoted())
return "\"" + sb.toString() + "\"";
return sb.toString();