The EDI file has a special character � (0xfffd) as segment delimiterThough it is nowhere mentioned specifically but as a best practice, EDI delimiters should be single byte characters (with CRLF as exception). If you are choosing any multibyte character as EDI delimiter (including CRLF) then you are definitely going to end up in a tricky situation one day because EDI files are flat files and transferring them from one system to another may corrupt the file itself as different operating systems (with different encoding and character set conversion rule) may interpret multibyte characters differently.
If we change delimiter to ~, then B2B process successfully.As expected because tilde (~) is a single byte character (unicode 0X007E).
This same segment delimiter is processed fine in 184.108.40.206 but doesn’t get parsed in 220.127.116.11.I don't think that Oracle B2B ever supported any multibyte character as EDI delimiter and specially this character must not have been supported. If it ever worked with B2B then it might have been a bug (which is the least possible) or you are overlooking some details. If you can forward your PS2 (18.104.22.168) agreement export along with the sample payload which worked in your environment, to my id then I may quickly verify. You may also consider logging a SR with support if you need formal confirmation from Oracle on this.