Ok, this is not obviously broken for me. I need a bit more information from you. The Bulk Load snap uses the bcp utility to load data into SQL Server.
The data fed into the Bulk Load snap is written to a temp file using custom delimiters. Specifically these:
FIELD delimiter: [|!,!|]
ROW delimiter: [|!/n!|]
We invoke bcp specifying the temp data file plus the custom delimiters. The data file would of course explain what we’re seeing in the db, though this is a bit tricky to inspect because it is deleted after the load snap finishes.
I see a space after the first json block before another field delimiter, which should be a record delimiter, and then repeat of data. Hmmm…
What version of SQL Server are using? The version of BCP would be helpful too (bcp -v). Are you running a Windows or Linux plex? Sample input would be helpful – maybe you can download the data that is fed into the Bulk Load snap.
We can actually replay the data and invoke BCP from the command line. It’s good you checked that could load data using SQL Insert – a nice cross check. Something with the shape of the data going into bulk it seems.