On 29 March 2020, the number of unreviewed good article nominations reached 715, an all time high. This number was sufficiently high that the page listing good article nominations reached its maximum size, and could not transclude any more reviews, resulting in some nominations failing to be listed on the page. BlueMoonset started a discussion on the GAN talk page, and as a result a backlog drive was begun that has brought the number of unreviewed nominations down to 232, as of publication.

As long as there have been good article nominations, reviewing backlogs have been a concern. In 2007, the Signpost noted rather drearily that "backlogs continue to grow", and by May a relatively informal backlog drive was in progress. The drive ended in June and another was quickly begun in July. That drive alone claimed to have resulted in 406 good article reviews, driving the backlog down to 82 outstanding and 54 unreviewed nominations. A third drive that year, planned for September, was cancelled after it was determined "more time was needed to avoid reviewers suffering from burn out". Ten subsequent drives from October 2008 to August 2016 saw varying amounts of success.

After a three year break, another drive was organized in September 2019, with Barkeep49 and Lee Vilenski acting as coordinators. On 1 September the backlog hit 626 nominations and 533 unreviewed. By 1 October, the totals were down to 463 and 337, respectively. As with all drives, after its success, the backlog again began to climb. By 1 February 2020 it had reached 629, surpassing the total before the drive. Wikipedia:WikiProject Good articles/GAN Backlog Drives

Graphs are unavailable due to technical issues. There is more info on Phabricator and on MediaWiki.org.

As the number of outstanding nominations peaked at 715 on 29 March, a backlog drive set for April and May was organized by Harrias. Harrias and Lee Vilenski agreed to serve as coordinators.

After beginning on 1 April, the number of outstanding nominations was already down to 693. The number of unreviewed nominations dropped from 605 to 519, or a 14.9% change in only two days. In the first six days, the number of unreviewed nominations dropped by more than twenty per day, peaking at 50 on 3 April. It was then above ten every day until 16 April. Throughout April the drive saw general success – as of 22 April, there were only 426 outstanding nominations and 243 unreviewed, a drop of 362 articles, or 59.8%.

As of 27 May, 1079 good article reviews had been conducted by 147 users, bringing the backlog of unreviewed nominations down to 215 (a drop of 64.5%) and the number of outstanding nominations to 340.

A staggering nineteen users have completed at least 15 reviews, fourteen at least 20, eight at least 40, and five over 50. The Rambling Man, who alone conducted over 100 reviews, considers that "the drive has taken advantage of the pandemic, many of us have more time on our hands to get cracking with reviews and that it coincides with a massive backlog at GAN was somewhat fortuitous."

Lee Vilenski and Harrias spoke to the Signpost about the success of the drive:

I think any area that promotes a standard for how articles should be written, and rewards articles that are well put together should be lauded. Reviewing said articles, however, is often seen as a form of QPQ or as time that could else be used to work on other articles. This backlog drive's success is due to a community spirit ethic to shorten the queue, and also a competitive attitude to do as many within the two months as possible. So far, there have been users who have completed more reviews in a month, than I have in two and a half years! A quasi-competitive and team dynamic to reduce the backlog has all contributed to the current success.
— Lee Vilenski

Honestly, the main answer to this is "Covid-19". But even accounting for that, there has been a massive buy-in: people could just as easily be using this time to write or expand articles, but we've had a brilliant level of participation. I think the high level of the backlog will have been a factor for some: it got so big that the transclusions were actually exceeding the maximum limit, so some nominations weren't even appearing on the GAN page. Most volunteers (and we're all volunteers on Wikipedia, after all) tend to want to help make things better, so when people see a problem, they want to help solve it.
— Harrias

The good article project no-longer seems to have formal leadership to ensure quality. The associated WikiProject has been tagged as inactive, meaning that the very project that "is designed to maintain the Good articles list and oversee all other GA related tasks" is not carrying out its function. There is no way of ensuring uniform quality reviews other then editors checking each other's work. Though GA sweeps were conducted in 2010, 31% of all good articles currently have a cleanup tag.

What solutions will work to keep the backlog manageable in the future is "The million dollar question!" Harrias wrote:

To me, the ideal is better WikiProject buy-in: you and I are both active in the MilHist community, and that tends to have a relatively low backlog. But that ship seems to have sailed; for the most part, the WikiProjects are all but dead. In the medium-term, I would like to have quarterly backlog drives: a one month drive, followed by a two month break. But that remains a tactic that does like more than paper over the cracks. A long-term solution? I really don't have a clue.
— Harrias

Though backlog drives serve well to keep the number of nominations manageable, the general trend is that the number of nominations goes up quickly after a drive is concluded.

The main problem with GAN is that reviewing is highly concentrated among a small number of users, while nominating is much more broad-based. (If you look at WP:GAN, you can see that the vast majority of reviewers have done more reviews than the nominator.) Many nominators are new to the process and others are prolific nominators uninterested in reviewing. Requiring a quid pro quo, as was instituted at DYK, would help expand the base of reviewers, but the cost would be more pro forma reviews and pushing away would-be nominators who write quality articles. (Restricting the number of nominations wouldn't help much. Even looking pre-backlog, only a small fraction of nominations are above the suggested threshhold of five nominations per user. Perhaps this is because nominators with a track record are more appealing to reviewers, leading to shorter waits for them.) The backlog drive seems mostly to encourage increased participation in GA reviewing by people who were already prolific reviewers, and it's equally concentrated; the top five reviewers or so have done a majority of backlog reviews to date. On the other hand, the drive has significantly increased overall reviewing and shrunk the backlog
— Buidhe

The only thing that has become abundantly clear is that something needs to change to allow the GA project to function efficiently and effectively. Steps to take may include:

1. Reactivation of the associated WikiProject, which was marked as inactive on 21 February. Many users are already active in GA-space, and will serve as a starting point for larger change.
2. Conduction of broad GA sweeps. While there's not enough energy to review every single good article, reviewers should check articles with long standing, major cleanup tags, put them up for a brief review, and have authority to demote them.
3. Appointment/elections of co-ordinators for the project. While several users function unofficially in this capacity, it will be useful to have several users who have the responsibility for following up on stale reviews, making sure every question gets a prompt answer, guiding editors, spot-checking new reviews, and spearheading change. This doesn't have to be a terribly time consuming role, especially if the responsibility is spread out among several users.
4. A new bot. Legobot has not been edited for a long time, and a new bot could incorporate new features. Lee Vilenski spoke further on the topic: "Legobot has done a fantastic role for us for many years, but the lack of changes that can be made have made the need for a new bot a necessity. This would allow us to further split down our nominations into subtopics stating where our backlog is, and put additional work into. I'd also suggest ratio of nominations/reviews shown against any new nomination rather than just reviews as is done currently."
5. A newsletter to keep interested users up to date on potential drives and the state of GA.

But above all, we should gain an atmosphere where doing reviews is highly positive. There are many users who find the review process to be daunting or overly critical; whereas they could be the next serial reviewer. Having a positive place to quality assess reviews and offer constructive help is the way forward, rather than chastise a review for being of poor quality. The quality of Good Articles has never been so high, so I'd like to congratulate everyone on such hard work!
— Lee Vilenski