Comment 32 for bug 1704396

I was able to recreate a few searches that showed the delay between the "Firing off the multiclass query" / "Returned from the multiclass query" pair as Mike mentioned in comment 28, but have not seen any instances of "Error printing XMPP stanza to socket" in the logs.

An observation I found interesting:

Two metarecord keyword searches for "apple tree" both showed the delay before returning results. The timelogs showed the delay in different places.

For the first search, the delay was between the FIRST "Firing off the multiclass query" / "Returned from the multiclass query" pair.

For the second search, the delay was between the SECOND "Firing off the multiclass query" / "Returned from the multiclass query" pair.

Viewing the search results where the time lag showed between the FIRST multiclass query pair, the first record in the list of hits showed the following:

- no record id in the title link
- no format icon and label

Viewing the search results where the time lag showed between the SECOND multiclass query pair, the second record in the list of hits showed both the missing record id and the missing format information.

Not sure if this is really new information or just restating what's already been reported, but might this suggest where more logging could be inserted to help narrow this particular facet of the problem?