Most tabletop campaigns that fall apart do it around session three. Not session one, when everyone is excited, and not session ten, when a real group hits a normal rough patch. Session three is when the novelty wears off and the mismatches that were there from the start finally surface. The collapse is rarely about one big fight. It is four quiet mismatches that were baked in before anyone rolled a die.

The honeymoon hides the cracks

Session one is a bad test. Everyone is on their best behavior, the premise is fresh, and nobody has had to reschedule yet. The problems that end campaigns are structural, and structure does not show up until the structure is tested. That first test is usually the third or fourth time you try to get five adults in the same place at the same time.

The four mismatches

Schedule

The quietest killer. Two players can only do Saturdays, the GM only runs Wednesdays, and everyone agreed to "figure it out" in session one. By session three the calendar is a negotiation, two people are always missing, and a missing-player session feels like a waste, so the next one gets cancelled too. Scheduling is not a detail. It is the load-bearing wall. We wrote more about that in why scheduling is the hardest part of D&D.

Expectations

Someone signed up for a heroic story and someone else wanted a gritty survival grind. Both are great games. They are not the same game, and the gap becomes obvious right about the time the plot asks the party to make a hard choice.

Play style

Combat-heavy versus roleplay-heavy. Sandbox versus a guided story. Rules-strict versus rules-light. A table that is split on these does not argue in session one. It quietly disengages by session three, with one player on their phone and another wondering why this is not fun.

Reliability

Two cancellations in three weeks and a campaign starts to dissolve, because momentum is the only thing holding a recurring commitment together. Once "we'll skip this week" becomes normal, the campaign is usually already over; it just has not been announced.

How you actually prevent it

You cannot personality-test your way to a perfect group, but you can stop putting obvious mismatches at the same table. That means screening for the four axes before the campaign starts: confirm the schedule really works for everyone, align on expectations and play style out loud (a Session Zero is built for exactly this), and pay attention to reliability signals.

This is the entire reason Guildhall scores compatibility before you apply instead of after. A table that fails on schedule or boundaries never reaches your default view, so the session-three collapse you are trying to avoid is the thing the matching is designed to prevent. The right table starts with the right four people, not the right plot.