* several different schedules for a job
separate file? separate directory with schedules?

* schedule type: once, once and delete afterwards, at startup
execution_interval, execution_schedule, execution_type (once, once && delete, at startup)

* schedule format: configurable time zone for each job
(makes sense only for execution_schedule/cron type jobs?)

* not schedule jobs on holidays
definition of holidays -- list of dates? sync with what?

* move holiday jobs to other days
last_run? (state retention)

* change schedule a bit to not run some jobs in case of emergency
disabled: disable all, enable all, disable list, enable list?
(implemented in yandex version, to be pushed to public)

* chain of jobs: if one fails, stop and alert
each has its own schedule? (state retention)

* error recovery: if job fails do something (depending on the type of error?)

* test run job by name, with snaked environment
(foreground? logging?)