Max much slower when a M4L device is in edit mode.

I notice that Max for Live devices are overly slow when they are in edit mode, even when they are still frozen.

I see this with devices that iterate through the LOM, such as getting all variation and macro information of a rack device or, on a bigger scale, all devices in a Live set. Using the device can take up to a minute (including beach balling), whereas the closed devices do things in a few seconds. The bigger the Live set, the relatively slower an edited Max device is.

Testing during development can become quite tedious when the time differences are significant because I always need to store and close the device. However, even with small differences, I see things happening in Live's UI that I won't see when the device is closed, as everything happens quickly. Before I realized the device in edit mode was the cause, I was looking for solutions to solve this.

It happens in Live 11 and 12. Max 9.0 seems a tiny bit faster than 8.5/8/6.

Is it normal for an edited device to be much slower? If so, what's the reason?

And why do edited Max for Live devices have an extra latency of 4096 samples? Is this related to the slow speed?