Networking was Easier than the Game Menu
Heave Ho's ragdoll physics were already stretched to their limit just running locally on the original Nintendo Switch. Floppy characters, grabby hands, a rope system tuned right up to the edge of what the hardware could handle. When Le Cartel, the studio behind Heave Ho 2, and formerly Mother Russia Bleeds, started asking around about adding online multiplayer, the answer was the same every time.
"We had talks with other studios and they all told us: the game relies strongly on physics. It's four players. It's really real time. You are a small team. It will be complex," said Guillaume Levieux, the programmer behind Heave Ho 2. "We were kind of convinced it was not possible."

A three-person studio
Le Cartel is small on purpose. Guillaume handles all of the programming. Frédéric Coispeau is the Creative Director. Alexandra Muttoni is the Art Director. That's it. That’s the studio.
Heave Ho, now closing in on 1 million copies sold on PC and Nintendo Switch, is the result of a 48-hour game jam, the kind of prototype that wasn't supposed to go anywhere. Le Cartel had a meeting booked with Devolver Digital to pitch a different game, one they'd been building for six months. They showed the jam prototype instead. Devolver signed Heave Ho and the six-month game went into a drawer. It shipped in 2019, just ahead of COVID, and found its audience on couches and on streams: a game about grabbing your friends' hands and swinging across platforms while everything wobbles.
"We were a small team, there was only one programmer," Guillaume said. "We thought online play would be technically very complicated. Heave Ho shines when you're playing with friends on the couch. So we gave up the idea and just focused on local multiplayer."
The alpha problem
By the time Heave Ho 2 hit alpha, roughly a year and a half into production, Devolver had seen enough to know the content was there. What was missing was a reason for the sequel to be more than "a very good 1.5." Streaming culture had moved on from "buy a console and share a couch." Online was the obvious differentiator. It was also the thing Le Cartel had already decided, once, wasn't possible for them.
That's when Devolver's CTO John Bartkiw got involved. He knew a team called coherence who'd do a proof of concept for free. If it worked, they'd talk. If it didn't, at least everyone would know for sure.
Le Cartel was not optimistic, going in. Every developer they'd asked had told them the physics made this a bad bet.

The kitchen level
Before any contract existed, coherence ran two proofs of concept. Cary Brisebois (The Simpsons: Hit & Run, Crash Bandicoot - Crash of the Titans) built the first one alongside Benito Rodríguez (Crysis 1, 2, 3, Ryse: Son of Rome), proving the basic idea wasn't dead on arrival. Benito then carried it further alone for a second round, and for that one, Le Cartel didn't hand over an easy level to make the technology look good. They handed over the hardest one they had: The Kitchen. It had ropes, cooking equipment, ingredients flying between players, and physics interactions stacked on top of each other.
"We chose it because you have many interacting objects, and the most complex objects in the game are in the cooking levels," Guillaume said. If this level worked online, the reasoning went, anything would.
Guillaume showed up to the first online playtest expecting the worst, and for his sceptical colleagues to be proven right.
"We knew it was going to be a bit laggy. You can't have everything. If you want accuracy, you need to accept some lag. We were really just looking at: is it playable? Can we actually play this?"
It was playable. "When I played it, I just thought, oh, okay, my character is maybe a little heavier than I feel locally. But… it's okay! I cannot describe the relief I felt."
John Bartkiw, the CTO of Devolver, had a go as well. It was his first time ever trying Heave Ho online, and he was convinced. The proof of concept surfaced exactly one real problem: a visible lag between a character's hand and the rope it was holding. Le Cartel flagged it, coherence explained what fixing it would take, and that was enough. Full production was greenlit, with Benito staying on as the engineer who'd carry the project from prototype to launch.

Heavy rope action, and how to network it
Heave Ho 2 leans on ropes constantly. Some levels have fifty or sixty of them, each with up to dozens of physics nodes. Syncing every node across four players at full rate would mean an unworkable amount of traffic. Cary had the initial idea for getting around this: don't sync every node, sync a fraction of them, and let the physics engine interpolate the rest locally. Ropes nobody's touching sync at a low rate; the moment a player grabs one, it jumps to full rate. On top of that, position and rotation data got compressed to fit the game's 2D reality instead of carrying a full 3D transform, cutting the data per node by roughly 60%.
A physics-heavy game, it turns out, doesn't necessarily have to be a bandwidth-heavy one. You just have to know when to let the simulation do the work instead of fighting it.
How a three-person studio and one support engineer actually worked
coherence's usual process runs on branches, pull requests, and code review. “The usual” for game development. But Le Cartel doesn't work that way. Everyone pushes straight to main. Yeah, it’s kind of nuts.
"These guys don't have any of that," Benito said. "They just work on a main branch and everybody pushes to main. But they are three people, right? And they were very close." He kept one branch open, once, for a large rewrite Guillaume needed to review properly. Everything else went straight into main, in small, frequent pushes rather than anything held back for days.
The two teams settled into a rhythm of staying out of each other's way. "We tried to only work on things when they were done," Benito said. "They would tell me, okay, we're done with this adventure, the cooking, we're not going to touch it now, you're free to go in."
Benito also built a fast local testing loop, using Unity's two-window multiplayer play mode, so Le Cartel could test a scene in multiplayer the same way they already tested it solo. Later in the project, once that workflow was solid, Benito handed multiplayer testing back to Le Cartel entirely, backed by a written playbook, so the team could self-check changes and only loop him in when something needed a second look.
Daily updates went out in Slack. Guillaume didn't wait on formal review cycles when he had a question. "We could always ask. It was much better to have Benito available when we have issues. I just DM a question and I have feedback very quickly." Sloan (X-men: Destiny), a second coherence engineer, joined later once Le Cartel's design added more modes than originally scoped.
"Big shout out to Benito,” Guillaume said, “You’re very lucky to have him. He is very, very good. He's always making things perfect."

The game menu that took longer than the ropes
Ask Benito what actually ate the most time on this project, and it isn't the physics.
"One of the things I spent the most time on is the menu. Surely one to two months of work."
The reason comes down to a design decision Le Cartel wasn't willing to compromise on. Most online games treat the menu as a waiting room: everyone sits in their own copy of it until the host starts a match, and the group can dissolve the moment it ends. Le Cartel wanted the couch experience instead, where everyone is looking at the same screen together the entire time.
"What they wanted is every other player to navigate to follow the navigation of the host," Benito said. "It's like a remote control menu. I had to implement this for every menu screen. The host scrolls through the levels and people see it scrolling on their own screen too, while at the same time locally they can still do their own things."
The unlock system added another layer, and it got reworked more than once before landing. Levels unlock as a shared pool across the party, so a newer player gets access to whatever the group has cleared. Characters stay personal, you play what you've unlocked, not what the host has. Achievements are personal too. And the session itself doesn't reset between matches: a group stays together through adventure levels, then versus rounds, then back to the menu, all as one continuous session, the same people the whole way through.
"It's like replicating the local experience of playing locally together, but online," Benito said.
On Le Cartel's side, this was as much a design problem as an engineering one. "We had to revisit the macro design," Guillaume said. "It was not clear when we validated the prototype how Heave Ho 2 would be online. It was chaos. We looked at other games, like Overcooked, for inspiration on the flow." Two months of work later, it wasn't chaos anymore.

A new, coherence-inspired game mode
The clearest sign that coherence’s technology changed what Le Cartel could build is a mode they can barely test yet: Shuffle, where you play with a complete stranger matched online, no voice chat, just the game. It's the exact opposite of everything Heave Ho was built around.
"It even made us make something we wouldn't have made if we wouldn't have coherence," Guillaume said. Steam Remote Play, or any couch-sharing workaround, couldn't produce this. Matching with a stranger anywhere on the internet needs to be built into the game itself.
There's no way to fully test it before launch. "There are no players in the game for now, we don't know how it's going to behave," Guillaume said. He sounds curious about that more than worried.
Heave Ho 2 is available on PC, Nintendo Switch, and Nintendo Switch 2.
Try & buy Heave Ho 2 on Steam. Follow Le Cartel on Instagram and LinkedIn.
coherence is a multiplayer networking engine and co-development studio built by a team of veterans from DICE, CD Projekt Red, Crytek, Ubisoft, Blizzard, and Playdead. Studios use coherence to add or build online multiplayer in Unity without rewriting their games. For more, visit coherence.io.
Written By