For Sewers Tilesets (Act 3/Act 8) you can go to PlayerMovers -> OldPlayerMover -> MoveRange, and try to increase that to like 20ish and it should help alleviate the problem. I don't know if that would cause any side effects in other tilesets though, so that's why there was no change for 3.3.
The problem is a client issue. When the move is moving forward with the cursor in front of it, the game maps the actions behind you, or into walls or ceilings, resulting in the bot getting stuck. The only fix is to try and write a comprehensive playermover that dynamically adjusts the move range for these areas, which is on the todo list, but for now you can help the bot out, or use the new player mover setup to adjust your settings. If you play by hand a replicate the movement, you'll notice the same exact problem.
The old City of Sarn roof issue, which I think Quay is a similar tileset is the same exact problem, but those issues also are more just the nature of A* pathfinding and shortest paths. When you path towards a target, it tends to hug walls since that's how it gets to the destination the fastest. The solution to avoiding that is to do post processing n the path and manually move all points further away from walls, which there's some basic code for that, but it's something that you have to take a lot of time to test and change to make sure you don't create unintended side effects.
With how much the terrain changes over time and layouts change, given the game is procedurally generated, these are not easy problems to solve, and why certain areas just have problems and most of the game is otherwise good enough. Basically, we need to write area specific code, but that also means reresting all changes to make sure we don't introduce new issues, so for now, the areas that have issues have long been known and talked about in Release posts, but it's always easier to just say the game has changed terrain generation again, watch for stucks and help your bot.
Hopefully that makes sense.
