OK, something is wrong
I removed the retention file and re-scheduled a clean downtime, still the same behavior, and this in the retention file :
Code:
('in_scheduled_downtime_during_last_check', False),
('pending_flex_downtime', 0),
('downtimes', [<shinken.downtime.Downtime instance at 0x18ef7a0>]),
('in_scheduled_downtime', False),
('scheduled_downtime_depth', 0)
Note the 'in_scheduled_downtime'=False.
The downtime was scheduled with the current date + 2h, flexible.