After 49 hours of silence, Masangsoft is updating the "Agent" servers — and the game is running again

Masangsoft finally spoke. At 11:07 UTC on Wednesday, after 49 hours without publishing a line, a notice went up on Steam about the failure that had players reaching the room and being unable to play.

And roughly an hour and a half later, the game was working again.

The notice, in full

It is short — this is the entire text:

“Greetings from the GunZ Online Team.

Agent server updates will be rolled out sequentially to improve network stability, starting with Europe. Please note that during this process, some ongoing matches may end and players may be returned to the lobby.

We sincerely apologize for any inconvenience this may cause and kindly ask for your patience and understanding. Thank you.” — [Notice] Connection Issue, Steam, Aug 26 at 11:07 UTC

The post is pinned to the top of the game’s news hub — the company prioritised it.

What it says, and what it does not

What it says: there is corrective work under way, it is on the Agent servers, it is staged by region, and it starts with Europe. During the operation, matches may drop and return players to the lobby.

What it does not say — which matters just as much:

• it does not use the words “maintenance”, “hotfix”, “fixed” or “resolved”;
• it does not name any of the symptoms: no mention of 0 ping, of the connection icon over players’ heads, or of invisible players;
• it does not say when the other regions will be covered. No North America, no South America, no Brazil;
• it does not mention crashes;
• it says nothing about compensation. None of the game’s 25 official announcements to date uses that word;
• it does not open a maintenance window. Masangsoft’s maintenance notices follow a fixed format, with date, downtime, duration and purpose blocks. This one has none of them.

One more point of precision: what is new about this post is chronological, not thematic. Producer PD Max’s August 21 letter already carried a section literally titled “Connection and Crash Issues”, calling it an absolute top priority. What changed today is that, for the first time in this wave, a notice points at a concrete action and a specific layer.

So: an implicit acknowledgement of network instability plus the announcement of a fix in progress. Not a declaration that the problem is over.

“Agent”: the word that was missing

Yesterday we recorded here that the phrase agent error — the historical name for this failure in GunZ — appeared in none of the player threads in this wave, and that the word “agent” was absent from every official announcement.

Both flipped, and the source flipped with them: it was Masangsoft that introduced the term. It is now the operator, not the community, pointing at the “Agent” layer as the target of the fix.

As for what that layer is, a caveat is owed: in GunZ’s classic architecture, as documented by the private-server community, the MatchAgent is an intermediate component — clients connect to it while it talks to the central match server. That is consistent with the symptom we saw (the room exists, but the connection between players never closes), but it is legacy-code documentation written by third parties, and the 2026 client was rewritten. There is no way to claim today’s Agent works like the 2009 one.

What can be said is simpler and stronger: the company is working on exactly the layer the community had been pointing at.

Is it back? Partly — and the window can be pinned down

This is where the reporting earns its keep, because nobody, the company included, said what time service returned. What we have is a tight window built from verifiable markers:

• 11:53 UTC — forty-five minutes after the notice, a player was still writing on the forum: “it’s been 2 days an no solution still”. Another agreed in the same minute.
• 12:40 UTC — our own newsroom got into a room and played normally, seeing the other players.
• 12:51 UTC — player Tefo wrote “FINALLY I CAN PLAY” and repeated it in two separate threads, eleven seconds apart.
• 12:53 UTC — the concurrent player count captured by our monitor jumped from 614 to 662 in nine minutes. By 13:14 UTC it was 691.

That last marker counts for more than it looks: Tefo is one of the wave’s most persistent complainers. Over the previous twenty hours he had posted in at least four threads saying he could not play since the patch. The person celebrating is exactly the person who was stuck.

Note how tight the window is: only eleven minutes separate our newsroom playing again from Tefo’s public confirmation.

But hold the celebration. As of publishing, he was the only player publicly confirming the return, against more than thirty distinct accounts that reported the problem in the previous 24 hours. No thread has been marked resolved — and silence is not a fix: the people still broken may simply have stopped posting. And the notice starts with Europe, without naming Brazil, so we cannot say it explains our own recovery. That is timing correlation, not proven cause.

On our side, the reports logged on the status page also went to zero: none in the thirty minutes before publishing.

There is one indirect signal, recorded as a signal and not as proof: the top of the forum, which yesterday was nothing but connection threads, is back to being about the game — requests for more quests, K-Style discussion, weapon rentals, Australian servers.

The price of two days broken

The numbers show the damage.

Wednesday’s early hours produced the worst trough of the entire week: at 08:00 UTC the game had 482 concurrent players — the lowest figure across every point measured in the last seven days.

And the whole day sits below the previous one. Slot by slot against Tuesday, today runs roughly 20% lower across all seven measurements taken so far.

Base erosion, as a daily average of concurrent players:

Day	Average	Peak
Aug 22	1,057	1,241
Aug 23	1,032	1,153
Aug 24	803	1,047
Aug 25	786	979
Aug 26 (partial)	639	839

The game’s all-time peak, 2,241, has not been threatened since launch week.

The surprise: the score did not collapse

This is the counter-intuitive finding, and it is solid — we analysed the complete review corpus for the game on Steam, all 1,632 of them since launch.

The game sits at 1,632 reviews and 45.3% positive, a “Mixed” rating. Over the last 24 hours — the peak of the crisis — there were 36 reviews, 19 positive and 17 negative: 52.8% positive. That is above the game’s own historical average.

One likely explanation is that the complaint lives on the forum far more than in reviews. Across those 36 reviews, “crash” appears 3 times, “ping” twice, and “agent”, “VPN” and “BugTrap” zero times. Yet four of them describe the bug precisely, just without those words — like this one, in Portuguese:

“some matches you join and can’t play, all the players in the match just stand still”

And that crash window: what BugTrap is

A lot of people are running into a grey window reading “GunZ_Global — A crash has been detected by BugTrap”, with Submit Bug, More… and Close buttons. Worth explaining, because it looks alarming and is not what it seems.

BugTrap is not an error in your computer, not a virus, and nothing to do with Microsoft. It is a third-party crash-capture library for Windows, written by Maksim Pyatkovskiy and today distributed under an open licence. Masangsoft embedded that library in the GunZ client.

What it does: it replaces Windows’ default error handler and, when the game breaks, collects the fault address, the call stack and the state of the machine — log and minidump included. That window is the library’s default dialog, with not one line of text customised by Masangsoft. We checked the text against BugTrap’s original resource file: it matches word for word, three buttons included.

Why the title says “GunZ_Global”: BugTrap writes into the window the application name the game itself registered. “GunZ_Global” is therefore the internal name Masangsoft compiled into the client.

Is it worth clicking “Submit Bug”? Yes, for a concrete reason: the report goes to whatever destination the game’s own developer configured — their own server, HTTP, or a support email address. It does not go to Microsoft, nor to whoever wrote the library. It is literally the crash channel Masangsoft wired into the client. What cannot be guaranteed is that anyone is reading it on the other end.

This crash is old

An important point, so things do not get mixed up: the BugTrap crash did not start with this wave. We searched the Steam forums and it appears in three topics, all from the first week after launch — August 14, 17 and 20. There is no new thread about crashes or BugTrap today.

The profile players describe has been constant since then: the game freezes after about five minutes of a match, in deathmatch and team deathmatch, stuttering before it closes. As of Aug 22, nobody in the community had found a fix.

And there is a detail that grates: the August 24 patch — the one that, as we showed, carried not a single network item — did carry a crash item: “Improved stability to reduce occasional unexpected game crashes”. They touched it, and the reports continue.

No official announcement mentions the word BugTrap. Not one of the 25.

How to know when it is actually over

A practical tip worth more than refreshing the forum: Masangsoft usually does not publish a new post saying “resolved”. It edits the original post’s title and appends “(Resolved)”.

It has happened before: the “[Notice] Connection Issue” from August 15 is today titled “[Notice] Connection Issue (Resolved)”, its body rewritten to say the problem had been fixed.

So the signal to watch is exactly that: when today’s notice gains a “(Resolved)” at the end of its title, it is officially over.

Also worth recording: the official game site still has not said a word about any of this. The notice went out on Steam only; the official site’s board mirrored nothing, and its most recent post is about a content creator programme.

What to do now

1. If it is back for you, good — but it can drop again. The notice itself warns that ongoing matches may be ended during the rollout.
2. If it is not back yet, the rollout is staged. Europe is first in line and no timeframe has been published for the other regions.
3. Log whatever is happening to you on the status page. Besides the empty-room button there is a dedicated one for game crashes — and the chart now shows what the community is reporting, not just how many reports there are.

We will update this article when the notice gains its “(Resolved)” — or when things fall apart again.