Safari using significant energy

I've seen this ever since Safari 8 on OS X 10.9.5, OS X 10.10.x all versions with Safari 8 and sadly in Safari 9 on all El Capitan betas to date, too. The Safari memory leak in my case is severe and Safari has to be completely quit & restarted often. It only seems to happen if you tend to have a few windows open a lot, which you "reuse"; but overall, Safari just grows and grows (by many GB).

Suggestions about "putting in more memory" are absurd. I've a 16GB Macbook Pro laptop which is the maximum configuration of soldered-on ("pro" my rear-end!) memory which Apple provide. It simply isn't possible to add more. Memory pressure and slowdown tend to get critical when Safari exceeds 10GB. I did once persist to the point where it was using over 13GB. When restarted with all tabs manually revisited to ensure all pages are loaded, it'll go back to about 2.5Gb. A leak of that size is utterly indefensible.

This is a stark change in behaviour from Safari 7, which behaved basically fine in this regard - yet there are surprisingly few reports of it online. It isn't a subtle problem and Safari 8 has been around for ages. Others would have noticed, yet few report it.

I see it on my 10.9.5 machine, 10.10 home laptop, 10.11 test laptop and even, more recently, a 10.10 laptop at work. My conclusion is that this must be Safari screwing up when particular bookmark, cookie, cache and/or other data is present and this data must be part of the stuff it shares over iCloud - otherwise I would not have expected my independently clean-installed-by-IT-vendor work laptop to exhibit exactly the same behaviour.

Bottom line is that this seems to be a user data thing. Taking a deep breath and doing a complete Safari reset - ditching your iCloud bookmarks, emptying everything from every Safari instance on the iCloud account, deleting ~/Library/Safari and so-on - might work according to the Developer Forums. But as ever with Apple since roughly OS X 10.7, its a heisenbuggy mess and no amount of psuedorandom chicken slaying will be guaranteed to fix your issue.

Closed tab stray processes might just be down to a "bad extension", but that's no excuse - extensions are under Safari control, and a bad extension should never be able to break the browser. It's just JavaScript code executing completely under the browser's oversight. Still, we know that Safari must have very poor code for extension support given the problematic history, so that's always worth investigating if you haven't already.

TL;DR: If SystemUiServer causes about as many idle wakeups as kernel_task and you have iStatMenus, turn off all of iStat's widgets, then turn them back on. If you don't have iStatMenus, suspect other menubar widget applications. The SystemUiServer should have <10 idle wakeups.

I have another hardware-identical machine running El Capitan 10.11.6, and it doesn't suffer from excessive battery drain. When comparing the Activity Monitor CPU panes, ordered by Idle Wake Ups, the entries are similar, except that the SystemUiServer entry has very low wake-up counts on the El Capitan machine, and 100x higher counts on the problematic system - on par with kernel_task and in 300-500 range.

A search for issues related to SystemUiServer prompted investigation of widgets installed in the menu bar. On both machines, the additions over stock install are: iStatMenus 5.31, Caffeine 1.1.1 and OS X's Keychain lock.

Upon disabling all widgets from iStatMenus settings pane, the SystemUiServer became idle. Re-enabling the widgets doesn't bring the problem back. Further checks after a reboot and then after a shutdown-restart cycle show that the problem didn't come back.

The battery life is now normal. At idle, with screen at minimum brightness, and google.com open in Safari 10.0.3, coconutBattery 3.5.1 indicates a discharge rate of ~10W.