Please show/attach the SELECT statement and the record description.  (you can delete the data names for privacy, but we need to see the structure.)

Thanks Tom.

<<Please show/attach the SELECT statement and the record description.>>

 SELECT AUDIT
              ASSIGN TO DISC
              ORGANIZATION IS SEQUENTIAL
              ACCESS MODE IS SEQUENTIAL
              FILE STATUS IS FG4AUD-STATUS-CODE

FD AUDIT
      RECORD VARYING 0 TO 4096 CHARACTERS DEPENDING ON REC-LEN
      LABEL RECORD IS STANDARD

01 AUD--REC. |
    03 AUD--RECNAME   PIC X(12).
    03 AUD--TERMNO     PIC X(2).
    03 AUD--DATE           PIC 9(6) COMP-6.
    03 AUD--TIME            PIC 9(6) COMP-6.
    03 AUD--FUNC          PIC X.
    03 AUD--USER          PIC X(12).
    03 AUD--RELKEY      PIC 9(6) COMP-6.
    03 AUD--DATA            PIC X(4058).

Thanks Tom.

<<Please show/attach the SELECT statement and the record description.>>

 SELECT AUDIT
              ASSIGN TO DISC
              ORGANIZATION IS SEQUENTIAL
              ACCESS MODE IS SEQUENTIAL
              FILE STATUS IS FG4AUD-STATUS-CODE

FD AUDIT
      RECORD VARYING 0 TO 4096 CHARACTERS DEPENDING ON REC-LEN
      LABEL RECORD IS STANDARD

01 AUD--REC. |
    03 AUD--RECNAME   PIC X(12).
    03 AUD--TERMNO     PIC X(2).
    03 AUD--DATE           PIC 9(6) COMP-6.
    03 AUD--TIME            PIC 9(6) COMP-6.
    03 AUD--FUNC          PIC X.
    03 AUD--USER          PIC X(12).
    03 AUD--RELKEY      PIC 9(6) COMP-6.
    03 AUD--DATA            PIC X(4058).

OK, thank, Swapnil.

It is a binary sequential, variable record length file.

The structure of these files is:
       4 octet binary record length
       binary data for record
       4 octet binary record length

The record length is both before and after the data record; this allows both READ and READ PREVIOUS.

My guess is that, for this particular record , something has corrupted the record length counts so they do not match.  If the counts are not equal then the structure of the file is not correct for variable-length binary sequential ==> 98 errror.  Since the structure is unreliable, there is no way to read beyond the record using normal COBOL I/O.

Depending on how valuable this data might be, you can write a COBOL program to recover the data one byte at a time, using a select/file description of a fixed-length, binary file with record length = 1.  With a bit of COBOL logic, you can read through the file to check the binary record lengths (read 4 records, which determines the number of reads to pass the record, then read 4 more records, to compare the record lengths.  At some point, you will come to a record where the 4-octet record lengths do not match.  Depending on what you see, you can devise some means to pass the record, or rewrite (one byte at a time) the corrupted record length.

Good luck!