Collection data getting scrambled -- have you experienced this?

The data for one of our collections has become scrambled. On a community website, we have a directory of about sixty restaurants. It gets updated occasionally. Today we added a new restaurant and, when confirming how it looked on the directory page, we noticed that the directory suddenly had errors. We noticed some restaurants sorting out of place and some restaurants appearing twice.

We looked at the collection and discovered these problems:

  1. Some records had reverted to earlier versions. The most recent updates were missing.

  2. Some records were now duplicated, with both a current version and an earlier version.

  3. Some records had missing data in a field (hence the bad sorts), but that could be a special case of issue #1

We’ve got a call in to support for two reasons. Most importantly, we’d like to determine the underlying cause of the problem and know that it is being addressed. Secondly, if we can revert to a pre-scramble version of the data, we’d like to do that. Meanwhile, however, we’re working on manual fixes.

Has anyone else experienced this?

We have a believable explanation of what happened, although not what triggered it.

Support believes that a sync from a (long unused) sandbox…

  • Overwrote good data in some records
  • Created an extra record for others

This is believable. We did some comparisons and it looks like records that don’t exist in sandbox are unchanged in live, while changed live records were revised to match their sandbox counterparts.

So the issue now is what triggered this. The user who maintains the data does this on a regular basis and does not use the sandbox. It seems unlikely that she would have used the sandbox by accident.

This is an old site with one collection that uses the sandbox as part of the workflow. We’re going to redesign that workflow so we can completely disable sandbox for the site.