I think Figure 1 from Tietsche et al 2011 is important.
https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2010GL045698This shows two things - firstly, the general shape of the (modelled) decline in Arctic ice over the coming decades, and secondly the time taken to recover from extreme events such as the summers of 2007 and 2012.
For the latter, the conclusion is quite simple - the Arctic has a "memory" of about two years, and so any major excursion will bounce back to the long-term trendline within a couple of years. They only modelled downward excursions, but my guess is that it holds the other way too - even if by chance we have a particularly good year for ice retention, it'll be gone in another couple of years. The paper discusses the mechanisms for this, but fundamentally it's quite simple - if you have a massive loss of ice one autumn, that means a correspondingly massive extra heat loss in the following winter. By the end of spring, first year ice has grown back. A low summer minimum has very little effect on the following maximum. This is believable, and we've seen it after every major loss year for more than a decade now.
For the longer term decline, look at the
shape of the curve. Note how it's staggered and stepped. This reflects the shape of the Arctic ice basin. There are shallow seas around the edge, and a deep central portion that covers about 5 million square km. So, as ice loss progresses, there's an initial period of rapid decline that plateaus at around 4.5 to 5 million until about 2020. That's exactly where we are now, in that plateau, where the summer minimum roughly covers the deep parts of the Arctic Ocean but the peripheral seas melt out each summer. Subsequently, there's another period of rapid decline that plateaus again at 1.5-2 million. This is the "remnant above Greenland" stage. The final collapse comes after that.
The shape looks entirely plausible to me, all that we need to work out is the scaling on the X axis, and to be honest I'd be surprised if they're far off. Right now we're on the verge of the second period of decline - but it'll plateau again in another couple of decades, probably before hitting the "ice free" threshold of 1 million.
It may be we need to squash the X axis up by 10% or so to fit reality - someone with more time than I can probably make an overlay - but it's really not far off.