When testing out JSON functionality I can consistently repeat the following crash:

2021-08-17 08:13:16.563 EDT [37784] STATEMENT:  select imdb_id, title, imdb_rating, t.* from movies_json_generated_crash, jsonb_to_recordset(jsonb_column->'cast') as t(id text,name varchar(100),character text)  where name like 'Robert Downey%'

2021-08-17 08:13:16.564 EDT [37784] ERROR:  cannot call jsonb_to_recordset on a non-array

2021-08-17 08:13:16.564 EDT [37784] STATEMENT:  select imdb_id, title, imdb_rating, t.* from movies_json_generated, jsonb_to_recordset(jsonb_column->'cast') as t(id text,name varchar(100),character text)  where name like 'Robert Downey%'

2021-08-17 08:13:16.839 EDT [9849] LOG:  background worker "parallel worker" (PID 37785) was terminated by signal 11: Segmentation fault

2021-08-17 08:13:16.839 EDT [9849] LOG:  terminating any other active server processes

2021-08-17 08:13:16.839 EDT [37784] WARNING:  terminating connection because of crash of another server process

2021-08-17 08:13:16.839 EDT [37784] DETAIL:  The postmaster has commanded this server process to roll back the current transaction and exit, because another server process exited abnormally and possibly corrupted shared memory.

2021-08-17 08:13:16.839 EDT [37784] HINT:  In a moment you should be able to reconnect to the database and repeat your command.

2021-08-17 08:13:16.839 EDT [37640] WARNING:  terminating connection because of crash of another server process

2021-08-17 08:13:16.839 EDT [37640] DETAIL:  The postmaster has commanded this server process to roll back the current transaction and exit, because another server process exited abnormally and possibly corrupted shared memory.

2021-08-17 08:13:16.839 EDT [37640] HINT:  In a moment you should be able to reconnect to the database and repeat your command.

2021-08-17 08:13:16.839 EDT [37521] WARNING:  terminating connection because of crash of another server process

2021-08-17 08:13:16.839 EDT [37521] DETAIL:  The postmaster has commanded this server process to roll back the current transaction and exit, because another server process exited abnormally and possibly corrupted shared memory.

2021-08-17 08:13:16.839 EDT [37521] HINT:  In a moment you should be able to reconnect to the database and repeat your command.

2021-08-17 08:13:16.843 EDT [9849] LOG:  all server processes terminated; reinitializing

2021-08-17 08:13:16.884 EDT [37791] LOG:  database system was interrupted; last known up at 2021-08-17 08:01:20 EDT

2021-08-17 08:13:16.927 EDT [37791] LOG:  database system was not properly shut down; automatic recovery in progress

2021-08-17 08:13:16.928 EDT [37791] LOG:  redo starts at 26/263FFF30

2021-08-17 08:13:16.928 EDT [37791] LOG:  invalid record length at 26/263FFF68: wanted 24, got 0

2021-08-17 08:13:16.928 EDT [37791] LOG:  redo done at 26/263FFF30

2021-08-17 08:13:16.935 EDT [9849] LOG:  database system is ready to accept connections

*******************

Note this only happens with Parallel workers, setting max_parallel_workers_per_gather to 0 works around this.  

This happens when
A.) you are using the jsonb_from_recordset command
B.) Doing it across a large enough amount of data to use parallel workers
C.) There are differences in the records within the JSON Documents i.e.  

  "cast": null,

vs

  "cast": [

    {

      "id": "nm1865544",

      "name": "Stephanie Pearson",

      "character": "Ronnie Price"

    },

    {

      "id": "nm8692131",

      "name": "Hope Quattrocki",

      "character": "Dr. Jessica Barnes"

    },

This should generally throw the error:  

ERROR:  cannot call jsonb_to_recordset on a non-array

But will crash often on large datasets.