Steps to reproduce:

After thunderbird packaged was updated to version 91, I am suffering from the following rather annoying bug. When I am trying to read more carefully some thread and clicking on earlier viewed messages they may appear blank till restart.

The story with more details. A couple of months ago I selected a news group for offline use expecting to avoid network latency for fetching of messages I have read earlier. For some time I was happy but ubuntu package updated from version 78 to 91. New progress bar made fan in my laptop noisy (Bug #1753195). Another problem was that I could not save attachments any more Bug #1752806 comment #3, so I disabled offline storage. Now thunderbird can not display messages in certain time interval more than once. In response to display in separate tab both views become blank. Thunderbird can show source of such message or properly cite it for reply. Messages earlier saved in offline storage works reliably. More bizarre is that I have not noticed the problem with fresh enough messages even though they appeared after I disabled offline storage.

I hope, the following can be considered as a recipe to reproduce:

• Create a news account for news.gmane.io
• Subscribe to the gmane.emacs.orgmode group
• Fetch 18000 messages
• Close thunderbird
• Put the gmane.emacs.orgmode attached file to News/news.gmane.io/ in the profile directory pretending that offline storage were enabled some time ago. I squeezed it as much as I can and the magic completely disappears when I remove first or last message from remaining 16 entries.
• Read some oldest messages and try to click on just seen message again.

Actual results:

Message pane displayed blank, headers are not updated despite the message were successfully fetched and displayed short interval before.

Expected results:

I can read any news message as much times as I need.

mbox file, offline storage for the gmane.emacs.orgmode news group

How is 91.7.0?

Version 91.7.0 downloaded from https://www.thunderbird.net has exactly the same problem as 91.5.0 package from the Ubuntu-20.04 repository.

Is there any reason of hope that behavior might change? Latest non-ESR releases at least have another NNTP implementation, but they are affected by another bunch of bugs.

(In reply to max from comment #3)

...
Is there any reason of hope that behavior might change?

Only if the cause is found. The odds of that are greatly increased if you are able to find the regression range using https://mozilla.github.io/mozregression/

FWIW, the list of news bugs reported in the past year https://mzl.la/3CO0O3a

Latest non-ESR releases at least have another NNTP implementation, but they are affected by another bunch of bugs.

If you are able, please continue to file bug reports for those as well. Thanks.

(In reply to Wayne Mery (:wsmwk) from comment #4)

(In reply to max from comment #3)

...
Is there any reason of hope that behavior might change?

Only if the cause is found.

I faced this bug with 91.5 and I still have no idea why it might disappear in 91.7 without any efforts from the side of developers.

The odds of that are greatly increased if you are able to find the regression range using https://mozilla.github.io/mozregression/

I have quickly tried 68.9.0 and do not see the bug there with the steps suitable for 91. I am not going to continue partially because I believe that I provided enough info to reproduce the issue and so to debug it. The bug is quite peculiar and with my main profile and significantly larger file with offline storage it is more prominent, more messages are affected, so it is really painful. It is unclear in advance which messages will be unavailable. It is a waste of time to study a black box further when an example of buggy input is known.

It may me disappointing for you but I am disappointed as well. The 91 ESR version brought a lot of pain, so I am about to start trying alternative applications.

FWIW, the list of news bugs reported in the past year https://mzl.la/3CO0O3a

Sorry, but I do not follow. Most of the bugs are mine and have been filed during last couple of months. Originally I had an intention to file one new bug and one old bug that became more painful, but now I am feeling myself buried under an avalanche of problems. And last days I got the following wonderful feedback:

• attempt to reproduce in a completely unrelated scenario
• attempt to reproduce neglecting details provided in the later comments, and I am still really puzzled what is different in my "clean" environment that I have reproducible crashes
• request to retest with new version that resembles enterprise-grade support when your are complaining about a problem that appears at the end of a month and are repeatedly receiving notifications to retest in the beginning of next month.

If you are able, please continue to file bug reports for those as well. Thanks.

(In reply to max from comment #5)

(In reply to Wayne Mery (:wsmwk) from comment #4)

(In reply to max from comment #3)

...
Is there any reason of hope that behavior might change?

Only if the cause is found.

I faced this bug with 91.5 and I still have no idea why it might disappear in 91.7 without any efforts from the side of developers.

The odds of that are greatly increased if you are able to find the regression range using https://mozilla.github.io/mozregression/

I have quickly tried 68.9.0 and do not see the bug there with the steps suitable for 91.

What I mean is, I would not have expected a change between 68 and 91, because I would have not expected nntp to change in that period. But clearly something has.

I am not going to continue partially because I believe that I provided enough info to reproduce the issue and so to debug it.

My request was not a demand - I hope it didn't sound that way. It's just that a regression range by the people who can readily reproduce the issue does quickly identify the changed code. But we can proceede without it. Your steps are quite clear, so yes I expect that the cause can be discovered.

FWIW, the list of news bugs reported in the past year https://mzl.la/3CO0O3a

Sorry, but I do not follow.

I simply added it to correlate to other reported issues. That is all.

(In reply to Wayne Mery (:wsmwk) from comment #6)

(In reply to max from comment #5)

I have quickly tried 68.9.0 and do not see the bug there with the steps suitable for 91.

What I mean is, I would not have expected a change between 68 and 91, because I would have not expected nntp to change in that period. But clearly something has.

Some speculations. Conditions to reproduce are rather fragile. It might be a race, so behavior might be different due to some completely unrelated change.

Another case with similar behavior:
https://support.mozilla.org/questions/1369121

Notice that the "repair folder" recipe in the accepted answer has ugly consequences:

Thank you for taking the time to explain what "repair folder" does. Nonetheless, for reasons I don't understand, I did lose all the posts I had saved. I guess if I want them back, I'll have to restore them from a backup- it does no harm to try.

Actually I was checking if a bug for lost message states due to "repair" action already filed by someone else.

Max do you still see this when using version 102 or newer?