Reloading Map in Server Causes "Engine Error: failed to lock vertex buffer"

For the record, I'm running a GTX670, with an i5 and 16GB of RAM, on Windows 10.

Buckle up, because there's some weird shit going down, here.

____________

The Error in Question:

"Engine Error: failed to lock vertex buffer in CMeshDX8::LockIndexBuffer: nVertexCount=2204, nFirstVertex=0"

____________

Issue: I am building a new map, 'arena_mystic', and I've just about reached an early testing phase. Mostly blocked out with some basic texture work and a few models, with the game logic in place and working fine. No lighting work has been done, so it's using the global lighting setting.

• When I compile and run my map from Hammer, it works fine.
• When I run a new server from the button in TF2 using the map, it works fine *the first time*. Then, if I disconnect the server and return to the Main Menu, the *second* time it crashes.
• When running the server, if I attempt to switch to my map using the console command "map arena_mystic", it crashes. This is true regardless of whether the previous map was arena_mystic, or an official Valve map.

It's exhibiting some bloody weird behaviour.

____________

Other, Possibly Relevant Information:

1. Compile Log: Pastebin link.
2. Issue of Maximum Faces: One of the earlier issues I had with this map was that, due to some extremely enthusiastic work with the clipping tool, I had exceeded the maximum number of faces allowed by Hammer. I deleted these objects, and the problem seemed to go away. Likely not related, but potentially enlightening.
3. Issue with Model Clipping: Some of my prop_static models were clipping through my 2D skyboxes, and the compiler was complaining they were outside the level. This had the curious side effect of removing all other copies of that model, even those firmly inside the level. I deleted the clipping models as a temporary solution, the compiler stopped complaining about them, and the other models reappeared. Likely not related, but potentially enlightening.
4. The Saga of The GeForce Experience: When I first got this bug, I did the reasonable this and Googled it. Apparently the "CMeshDX8::LockIndexBuffer" bit is fairly common. Now, I know it's not an issue of too little memory, I have 16GBs of RAM. The alt-tab insufficient memory thing seems unlikely with a dedicated graphics card. That left the driver issue. I attempted to update it using the GeForce Experience program. GFE started to crash every time I booted it up. I uninstalled and reinstalled it. Now it doesn't even boot at all. I then updated the driver manually, by downloading version 353.62 from the Nvidia website and running the .exe file. That seemed to work, but has not fixed the issue. Almost certainly not related, but you never know.

____________

So, as you can see, this is what is technically known as 'a royal clusterfuck'. I'm sort of hoping that someone here has run into this before, and knows the cause. Any help would be much appreciated.

Nope

takabuschik said:

Um... you have a leak according to the compile log. Sort that out before continuing

Click to expand...

Found it, fixed it, still no dice. Tried cordoning off all the various areas too, still fucked.

Screw it, I'll just build a new map. This one's clearly irrevocably fucked.