The summary carries more weight than the sequence. Two of the four companies, the essay says, walked straight back to open licensing within two years [10]. Elastic's dual-license move and its AGPLv3 addition are forty-three months apart [1]. Redis is the case that fits: its gap is somewhere between five and sixteen months, and the source gives a year for the first move but not a month [2][6]. Forty-three months describes a retreat completed long after the fork it responded to had settled in, rather than a fear that evaporated quickly [3].
The label was worth going back for because packaging and procurement run on it. Debian, Red Hat and Fedora dropped MongoDB's package, and by 2021 the Open Source Initiative had said publicly that SSPL does not meet the Open Source Definition [2]. Elastic's stated purpose for the August change concedes exactly that dependency [7].
The essay also asserts one trigger every time: a hyperscaler ships a managed version of your project and captures the revenue you were counting on [11]. It names the competing service twice, DocumentDB against MongoDB [1] and OpenSearch against Elastic [3]. For HashiCorp it reports the company's own framing about protecting continued investment [4]. For Redis it names nothing [6]. Two of four [3] is a pattern worth knowing, though it does not function as a rule you can screen on.
What does screen is duller. In all four cases a single company changed the terms on code it controlled [1][3][4][6]. So a dependency review reduces to two answerable questions. Can one board change this license. Does that vendor's hosted revenue overlap a compatible managed service someone else could run. Neither answer depends on the essay's larger thesis being right.
Note where the reversals landed. AGPLv3 [7][8] is the license MongoDB left in October 2018 [1], so for two of the four the sequence ends at the third one's starting point, seventy-nine months later [4]. Being able to call it open source again is a claim about a label [7]. Whether AGPLv3 clears your own legal review is a separate question, and it is now scoped to a release rather than a project: Redis 8 and the version before it are different answers [8].
The author is not a neutral observer and does not pretend otherwise. He builds ZizkaDB under AGPLv3 from day one, deployed inside the customer's own cloud rather than his [12], and argues that the EU AI Act's push toward auditability and data residency favours that shape [13]. He also concedes the piece that would settle it: no clear evidence has emerged that the license changes improved revenue for any of the four [9]. A sample of four, with one outcome confounded by an acquisition [5], amounts to a pattern small enough to keep in your head rather than evidence about revenue.
