I don't have any magic solution for slow machines, but this problem also affects computers that should be able to run Glest just fine, and will, as long as no one assigns too many workers to the same resource patch.
The A* frequently 'floods' on path requests from harvesters, what should be short and quick path calculations instead involve searching for an unreachable destination, a node limit stops such searches eventually.
With the node limit exceeded, the closest cell to the target that was found to be reachable is selected and the unit heads there. This is the 'well masked' part, the behaviour is correct, the worker heads towards the resources, and will more than likely end up next to them and start harvesting, but what should have been a very short and cheap path calculation, was in fact quite the opposite.
Further, in a congested resource gathering area, that path will often be thrown away because it becomes blocked, causing the whole process to repeat. Get a few units doing this on a regular basis, and you have a lot fruitless path calculation going on.
I'll patch it up somehow, I've used a few different solutions for this problem, the latest is by far the best, but it's heavily reliant on a lot of code that I don't want to be moving into another codebase just now, as it may need some maintenance yet. I'll dig up one of the old solutions.
Cheers.