By Kelly Twigger
Podcast | Transcript
Everyone’s finally past the first question on hyperlinked documents: yes, they’re discoverable. That debate is over.
The real fight now is sharper—and it’s what makes the Carvana decision a must read: when does a party have to produce a contemporaneous (point-in-time) version of a hyperlinked document, and what does the requesting party have to show to get that relief? Because in the cloud, the “attachment” is living. If you produce the file as it exists today, you may be producing something that did not exist when the email was sent. And courts are increasingly unwilling to let the timeline get rewritten—whether intentionally or not.
The short version of what the court did (and why it matters)
In United Ass’n Nat’l Pension Fund v. Carvana, Magistrate Judge John Z. Boyle doesn’t treat “point-in-time” as a philosophical argument. He treats it like an operational problem that needs a workflow. And he builds one.
The Court recognizes what everyone in discovery already knows: hyperlinked documents create challenges that classic email attachments don’t. Attachments are fixed. Hyperlinks aren’t. But the Court also makes an equally important point: difficulty is not a withholding justification. If the linked document matters to the issues, you don’t get to opt out because it’s too much work.
So the Court strikes a balance the way courts increasingly do in 2026: it uses caps, selection, tools, and deadlines to force production in a way that is both proportional and usable. Plaintiffs can select 250 responsive emails from any of 25 custodians, and Defendants must use Forensic Email Collector (FEC) to retrieve and produce the most contemporaneous versions of any non-privileged hyperlinked documents tied to those emails—on a tight turnaround. That’s the Court saying: “We’re not going to debate whether this is hard. We’re going to make it manageable and then you’re going to do it.”
On the Google Vault side, the Court draws a line that matters for every team trying to get parent/child relationships back after collection: Defendants must produce responsive, non-privileged Vault documents, and then Plaintiffs can select a limited set—200 documents—for Defendants to attempt to locate a parent email if one exists. But the Court will not order a sweeping “reverse search” across email systems without evidence on how it would work and why it’s proportional.
And there’s your rule, in plain English: courts will compel contemporaneous versions, but they will not compel broad, burdensome workflows unless you prove you need them and prove they’re feasible.
What parties have to show to get “contemporaneous” versions
Here’s where teams get tripped up: courts aren’t treating point-in-time as automatic just because you say “we need it.” They’re treating it like any other discovery relief: you need a proportionality story and you need to show why the non-contemporaneous version won’t do.
In Carvana, the Court repeatedly ties the request back to a litigation need that judges understand instantly: “who knew what and when.” That’s the gateway. If the point-in-time version changes meaning—changes knowledge, changes the sequence, changes whether someone had the information at the moment decisions were made—then contemporaneous production becomes not just relevant, but necessary to keep the evidence usable.
But the Court also makes clear what doesn’t work: asking the Court to order a big technical lift without explaining the lift. When Plaintiffs essentially asked for a reverse-engineered search from Vault documents back into parent emails at scale, the Court declined—because Plaintiffs did not provide evidence about how the search would work, why it would be proportional, or why it would not be unduly burdensome. That’s a critical practice point: if you want the Court to order a complex process, you must come in with a plan, not a wish.
What producing parties have to show to resist it (and what won’t fly)
If you’re the producing party, “not feasible” isn’t a position—it’s a conclusion. Courts want the work behind it.
Carvana is part of the growing line of cases that signals this clearly: the difficulty of point-in-time retrieval doesn’t justify complete withholding. If a true “as-of” snapshot can’t be obtained, the answer is not “no production.” The answer is: produce the closest version you can obtain, explain what you did, and be prepared to support burden and feasibility with something more than generalities.
This is why you’re seeing courts impose sampling and caps. When the technology is messy, courts don’t let the mess decide the outcome. They reduce the volume, define the method, and require the parties to execute. That’s exactly what happens here with the 250-email selection process.
How this decision builds on the earlier Carvana orders
This January 12, 2026 order isn’t a one-off. It’s the Court taking what it already set in motion and turning it into a repeatable mechanism.
The August 21, 2025 order is the foundation: it’s where the Court established that hyperlinked documents would be collected and produced using reasonable efforts, including working through feasibility, burden, and the mechanics of producing point-in-time versions. (And we covered that August order on our eDiscovery Day webinar on collaboration tools, because it’s exactly the kind of “collaboration changed discovery” moment teams need to internalize.)
The November 19, 2025 order matters because it expands the custodian universe—so by the time the Court gets to January, it’s not dealing with a narrow pilot in theory. It’s dealing with a real custodial landscape in practice. The January order then does what courts do when parties can’t land the plane: it imposes structure and makes the workflow executable at scale.
The practical playbook (what to do before you end up briefing this)
If you want contemporaneous versions of hyperlinked documents, don’t wait to fight about it in a motion. Put the framework in your protocol and build a record early. Here’s what actually works:
- Define “contemporaneous” up front. Do you mean “as-of the email timestamp,” “nearest prior saved version,” or another proxy? Pick it. Write it.
- Explain why point-in-time matters in this case. Tie it to timeline, knowledge, decision points, reliance, intent—whatever the merits turn on.
- Propose a bounded workflow instead of demanding the moon. Courts love caps and sampling because it feels fair and measurable.
- If you want a broad workflow, bring proof. Show how it works, how long it will take, what it will cost, and why narrower alternatives won’t solve the problem.
- If you’re resisting, bring evidence, not adjectives. What did you test? What failed? What are the error rates? What’s the cost estimate? What’s your alternative proposal?
- Plan for Vault and linkage disputes. Vault docs without parent context are predictable; build a limited linkage process that the Court will view as proportional.
- Have a fallback when perfect point-in-time isn’t possible. Produce the closest version, with an explanation, rather than withholding.
That’s the lesson Carvana is really teaching: point-in-time production is not theoretical. It’s a managed, proportional workflow—and courts will require it when you can show it matters.
Minerva26 is the discovery strategy platform that connects case law, rules, and workflows so teams can turn decisions like this into a defensible plan.
If this helped, share it!
If you’ve got colleagues living in Google Workspace, Drive links, or Vault collections, send this to them—because this issue isn’t going away, and courts are getting more specific about what they expect.
- 🎧 Listen to Meet and Confer
- 📩 Get the newsletter
- 🔁 Share this post on LinkedIn
Case links:
