Pulse widths.
The norm that has been general for a lot of years is that neutral happens at 1.5mS, full forward at about 1.8mS and full reverse at about 1.2, with a requirement that avalid pulse must be more than about 0.7mS long. Some ESCs accept a longer pulse, longer than what they regard as "full forward".
Some go into a panic mode when presented with an overly long pulse and play safe by switching off.
The result there is that full forward is reached at about 3/4 stick throw, once the ESC is giving full power there is no way to turn it up to eleven, so it just gives full power until the signal gets regarded as invalid.
With a mechanical trim, the entire window is moved, so hitting the over length pulse scene is more likely. The slider just moves the body of the pot, causing the wiper to use different bits of the track. Using software control to trim is possibly the finest way of increasing confusion, especially if the ESC self programs.
The T5 mentioned sounds like it is capable of going out of the useful range. OK for servos to get more travel, less so for ESCs. Anything above 2mS is generally over length for a pulse to fit into the 50ppS window and still have room for the quiet period that is needed for the receiver to know that one pulse has finished and another is due. Might be different with digital systems, but the stuff that plugs into them still needs to work with the old "real time" gear.