I have this SQL statement:
SELECT * FROM converts
WHERE firstname.lastname@example.org' AND status!='1'
ORDER BY date ASC, priority DESC
You could quantize the 'date' ordering into 10 minute chunks, so how about ordering by floor(unix_timestamp(date)/600), and then by priority
SELECT * FROM converts WHERE email@example.com' AND status!='1' ORDER BY floor(unix_timestamp(date)/600) ASC, priority DESC
Though two dates can be still be less than 10 mins apart but straddle two different 10 minute "chunks". Maybe that is sufficient, but I think to do exactly what you request is better done by the application.
(OP requested expanded explanation....)
Take two times which straddle a ten minute boundary, like 9:09 and 9:11 today:
Suppose you had a higher priority row for 09:11 than 09:09 - it will still appear after the 09:09 row because it fell into the next 10 minute chunk, even though it was only 2 minutes different.
So this approach is an approximation, but doesn't solve the problem as originally stated.
The way you stated your problem, a high priority row could appear before one recorded several hours (or days, or months!) earlier, as long there was an unbroken series of lower priority row with an interval less than 10 minutes.