SQL Server CDC Mirroring – Internal System Error on Init/Resync

Hey, quick thoughts.

No active Fabric outage showing as of yesterday, so it's probably not a declared platform incident, though a targeted Mirroring regression isn't fully ruled out.

The first thing I'd check: Microsoft changed varchar handling on Nov 18, 2025. Tables created before that only support varchar(8000), and on restart a mirror may try to re-init under the new rules and choke. That fits your weird pattern. Running mirrors are fine, but stop/start blows up. Mirrored SQL Server also caps at 1 MB per column.

Second, check the tenant setting "Users can access data stored in OneLake with apps external to Fabric." If that got flipped off, the replicator can't reach replicating status and every table fails at once, which matches your Scenario 2.

Also worth confirming your SQL 2016 build is on a supported CU, since version-level support has tightened.

Then escalate to MS Support with the ArtifactId, mirror URL, failing table, and timestamp. Pull the Eventstream and monitoring logs first so you know whether it's failing at the replicator or a specific table.

Honest take: I can't say for certain which one it is. The varchar cutoff and the tenant setting are my best guesses because they explain "healthy until restarted" better than anything source-side, and you've already cleared the source side.
But the logs plus the ArtifactId are what will give you the real answer.

If this works for you, kindly mark it as the solution and give a thumbs up.

Best,
Shai Karmani

Let's connect in LinkedIn