Description of problem:
cockpit-session-recording uses the journal.journalctl method to retrieve a list of journal entries, process them, then present a list of Recordings in the UI. Intermittently(about 50% of the time), only a subset of the total entries are returned to the .stream(data) handler. When this happens the .done() handler never is called. 

The code we use:

https://github.com/Scribery/cockpit-session-recording/blob/master/src/recordings.jsx#L323

More details provided at the upstream issue: https://github.com/cockpit-project/cockpit/issues/12580

This leads to cockpit-session-recording not playing back all of the recorded output. For example, a recording containing the command 'cat /usr/share/dict/linux.words', the journalctl method may only retrieve all words up to 'm', or 'n', or 'r'

I believe this is also evident in the cockpit Terminal page, running a command such as:

   # find /

This is related to BZ: https://bugzilla.redhat.com/show_bug.cgi?id=1694846

Version-Release number of selected component (if applicable):
Cockpit version: 199, 193
OS: Fedora 29, Fedora 30, RHEL8

How reproducible:
Often

Steps to Reproduce:
- dnf install tlog

- tlog-rec -w journal cat /usr/share/dict/linux.words

- In cockpit module code:

 let cmd = ["journalctl", "-q", "--lines=all", "_COMM=tlog-rec", "--output=json"];
 cockpit.spawn(cmd)
         .stream((data) => {
             console.log(data);   /* If data contains Zyzzogeton, all entries were returned */
             if (data.includes("Zyzzogeton")) {
                 console.log('Found string');
             }
         })
         .done(() => {
             console.log('Done!');
         });

- Can also be reproduced by installing cockpit-session-recording, and attempting to playback a recording with flooded output(cat /usr/share/dict/linux.words)

Actual Output:
Intermittently, not all of the output will be provided in .stream(data) and the .done() handler is never called