This code, calling RIGHT() on a 1024-char UTF8 blob:

    with q (s) as (
        select
            cast(
                cast('AAA' as char(1021)) || 'ZZZ'
            as blob sub_type text character set utf8
        )
        from rdb$database
    )
    select right(s, 3) from q

returns 'ZZZ' (OK)

If I make it a 1025-char blob:

    with q (s) as (
        select
            cast(
                cast('AAA' as char(1022)) || 'ZZZ'
            as blob sub_type text character set utf8
        )
        from rdb$database
    )
    select right(s, 3) from q

the result is

    335544321
    arithmetic exception, numeric overflow, or string truncation
    Cannot transliterate character between character sets

With CHAR type UTF8 strings, this error doesn't happen.

With single-byte charsets, RIGHT() works correctly with text blobs of any size (also > 32K).

LEFT() does not have this bug, and functions correctly with multi-byte charset text blobs of any size.