Investigating bug 1529872, we found the problem is related to tree._columnsDirty.
The problem is slightly tricky to reproduce, as it shows only in a new profile where the column ordinals have not already been restored (such as in automation). Column reordering do not work for that first run.

The way it works, in _ensureColumnOrder() tree._columnsDirty gets set to false even if this.columns were not set. And then blocked by that check from setting up the ordinals properly.

I'm not entirely sure exactly how this all hangs together, but it's clear that setting columnsDirty is causing the problem , and the Thunderbird automated test case will pass when this is removed.

I'm proposing to just drop the _columnsDirty property. Looks like an optimization attempt of some sort, but not a very useful one.

Remove the tree._columnsDirty property which is causing problems for column reordering in a fresh profile.

The try run I made looks ok to me https://treeherder.mozilla.org/#/jobs?repo=try&revision=1490092507d5231d43bac54300a7c830c4a0c0f2&selectedJob=236993855

Victor and Brian, can we please move this forward.

Yes, this looked fairly safe/reasonable. Here's a talos push to make sure we're not regressing anything, as _columnsDirty looked like it was a performance optimisation.

https://treeherder.mozilla.org/#/jobs?repo=try&revision=1960c30d4e413b5c7b7532e9cfcaed26cbf2566c

Comment 4 sounds fine to me

Apart from being partly busted, the rest looks green. So is this good to go?

Yes, I r+ed it in Phab. The patch needs proper author information before landing though. Magnus, care to update the diff or do you want me to land this manually?

Thanks, looks like I got the phab and lando problems sorted out.

https://hg.mozilla.org/mozilla-central/rev/b89fa4b887f0