Could you update to the latest version and see if it's fixed there? AFAIK there was some work done on this. Feel free to reopen the issue if you still experience any issues with the latest version.

Today I have updated to the latest version (3.9.1). And there are still same problems. The visitor_days_since_first of all users visiting today change to zero.

@tsteur How can I reopen this issue, I cant find the entry.

@diosmosis do you remember maybe details about it? I think you looked at it a while ago

We changed it to set the value to 0 if the visitor wasn't known. There might be an issue w/ the code that checks if a visitor is known or not. @chenqingan can you see what visitor_returning for those visits is set to?

Okay.Then I found more problems.
Today the value is still zero. I updated to 3.9.1 at May 21th. I afraid that the value will be only zero later.
Before I update to the latest, such problem is existing, but casual and random. The percentage of wrong value is about 1%.

Hmm, I'm not sure what to make of that. @chenqingan do you have access to web server logs for your matomo (ie, apache logs)? If you can find the log entry for the request to matomo.php for a specific problematic idvisitor/time, can you check the value of _idts? If it's non-zero, then the javascript tracker is working properly and matomo is setting the value incorrectly. If it's zero, then it means the javascript tracker is failing to send the proper value. In that case it would be helpful to look at a few of the user agents for visitors where this isn't working.

grep -cE '_idts=[^0-9]+' matomo-access_log
Computing result is 0.
And all the _idts have values. The values are timestamps.

@chenqingan can you email the access logs for a couple problematic visitors to hello@matomo.org (grep the file by visitorid/userid)? We can try and reproduce the issue.

@diosmosis I have sent an email to you.

@chenqingan thanks for sending the logs, I see the problem, new_visit=1 is used which is exposing a bug in the visitor recognition code. Will look into a fix.

@chenqingan would you be able to check if this change works for you: #14491 ?

@diosmosis Im afraid that I couldn't deploy this merge to my online edition. Because I have only one productive edition that is used to assure apps work normal. Sorry :)