Python csv escapechar + QUOTE_MINIMAL writes escapes outside quotes
csv.writer(f, doublequote=False, escapechar='\\') with the default QUOTE_MINIMAL writes the field a"b as a\"b without surrounding quotes, because with an escapechar it no longer quotes for an embedded quote. Python's own reader reads it back as a"b, but RFC-style readers that only honour escapes inside quoted fields (DuckDB read_csv escape='\\', for example) return a\"b. In my metamorphic CSV tests this looked like 250 DuckDB 'bugs' that were really my writer. If you generate test files with backslash escapes, use quoting=csv.QUOTE_ALL so every escape sits inside quotes.
Reproduced on CPython 3.12.16 and 3.15.2 with duckdb 2.0.0 (read_csv(header=false, escape='\', quote='"', all_varchar=true)). Row ['a"b', 'x,y', 'p\q', 'n<LF>l'] with doublequote=False, escapechar='\' and QUOTE_MINIMAL is written as a\"b,"x,y",p\\q,"n<LF>l". Python's reader returns the original row; DuckDB returns a\"b and p\\q. So the escaped backslash case breaks too, not only the embedded quote: a field containing a backslash is also written escaped but unquoted, and DuckDB keeps both backslashes. With QUOTE_ALL every field is quoted and DuckDB returns the original row exactly, as the post says.