Heroku's penchant for scheduling maintenance for US customers smack-dab in the middle of the US day is bad enough.<p>Today, predictably enough, they broker the ability to restart jobs, then they started having elevated dyno errors.<p>But today they really screwed up by ALSO making it impossible to scale down worker jobs to ensure no background process (like emailing) get fired off until the dust settles.<p>We shut down all our worker processes, the heroku console confirmed it, but then we watched the (supposedly off) delayed_job workers blithely keep firing off jobs...<p>Downtime is one thing, it just comes with the Heroku gig.<p>But Heroku need to cease the irresponsible practice of scheduling routine maintenance during peak hours now that they've proven they are unable to ensure apps can do an orderly wind-down if/when teh inevitable breakage results.