The old link lives on
Updating the calendar reaches the calendar. It does not reach the mail, the note, the forwarded invite or the bookmark where the old link actually lives.
A client is moving a recurring live group session from one video platform to another at the turn of the year. On the planning sheet this is one line: change the link in the calendar entries and send the update to everyone. In practice the link in the calendar is the one copy you can still edit. The others sit in old invitation mails, in an assistant's notes, in a partner company's own calendar system, in a browser bookmark. Those copies will keep working, or keep half-working, long after the calendar says otherwise. Most of the planning ended up being about those copies.
Every link you sent is a copy
An invitation looks like a reference to one event. It behaves like a copy. The moment it leaves, the link inside it is text in someone else's system, and you can only update it where that system still listens to you.
Editing an existing entry and sending the change to all participants works well for the people who accepted the original invitation in an ordinary calendar. You do not even need a new entry, which matters, because a new entry would leave the old one standing next to it. But in this case some participants receive their invitations through their employer, whose assistant re-issues them from a different calendar system. The entry those participants see was never ours. Our update reaches the assistant, and whether it reaches the participant depends on a process at a company we have never spoken to.
Then there is everything that is not a calendar at all. People copy the link into a note, keep the original mail, open it from their history. None of that is reachable.
The calendar entry is the one copy you can still edit. Plan for the others.
Make the old one fail loudly
If you cannot update a copy, you can at least decide what happens when someone uses it. There are two possible outcomes. Either the old link still opens a room, and the participant sits in it alone and wonders whether they are early, or the old link fails with a message that the meeting no longer exists.
The second one is better, even though it looks worse. A dead link is a signal. It sends the person back to their calendar or their inbox, where the current link is. An empty room is silence, and silence costs the participant the session and costs the organiser a complaint a week later.
So the old meetings get deleted on the platform we are leaving, deliberately, once the new dates are set. What we have not solved yet is the step after the failure: whether the old address can point to the new one instead of just failing. I would rather ship the loud failure now and add the redirect when we find a clean way to do it than wait for the elegant version.
Single dates, no series
The obvious way to set up a recurring session is a calendar series. We did not use one. A weekly series does not know about public holidays, and in a year with two or three of them landing on the session day, a series produces entries that have to be edited one by one anyway, each of which again sends a change to everyone.
Instead, each cohort's dates are calculated from its start date and created as single entries. Each can be moved without touching the others. It is more entries and fewer surprises.
There is one more ownership question underneath. The existing entries had been created in a personal calendar of the person who runs the sessions, not in the company account. Whoever owns the event owns the link and decides who can update it. Moving the entries into the company calendar comes before the platform switch, because otherwise the switch would be done in an account the organisation does not control.
The automation reads the link at send time
Reminder mails for these sessions are being automated: one two weeks ahead with a calendar file, one the day before with the meeting link. The link does not get copied into the automation when it is set up. It is read from the calendar entry every time a reminder goes out. Otherwise the automation would become one more copy of the old link, and a particularly persistent one.
And if an entry has no link on the new platform, the automation does not send the reminder. It sends me a warning instead, three weeks, two weeks and one day before the date. A reminder with a wrong link is worse than no reminder, because it looks official.
The go-live of the automation moved from autumn to January, to the date of the platform switch. Building it for the old platform first would have meant automating a state that disappears two months later. The remaining dates on the old platform are handled by hand.
What I check on any link migration
- List where the old link lives, not just where it is managed. Calendars, mails, partner systems, internal notes.
- Update existing entries instead of creating new ones, so the old entry cannot survive next to the new one.
- Make the old link fail visibly. An empty room is the worst outcome.
- Avoid series where exceptions are predictable. Single entries calculated from a start date are easier to move.
- Check who owns the events before you change them.
- Let every automation read the link at send time, and let it refuse to send without a valid one.
A migration is finished when the old link stops working, not when the new one starts.

