The "All Notes Off" button is an important feature of nearly all MIDI based hardware and software, simply because things like what happened to you do happen. When you press a key, it doesn't "play a note," it sends the instruction to turn on that note to the MIDI output device(s). When you release the key, it sends the instruction to turn off that note. Other instructions tell the output to change volume and/or voice, etc. All of the instructins get from the input to the output as strings of little bits, ones and zeros.

It's sort of the nature of ones and zeros to sometimes not get to the other end of the wire. If an entire "end note x" doesn't get there, you get the famous "forever note" that won't stop. If one or two of the ones or zeros "change their personality" on the way through the wire, the instruction that's intended may arrive at the output device as an entirely different kind of instruction. (Make me real loud, perhaps?).

In many cases, yours probably included, it's difficult to find where the "errant command(s)" came from. It's about as practical to assume that some fat zeros got stuck in the wire as to look for a "more technical" answer. (Jaggy ones, with sharp edges that hang up are another possibility. You could also presume that the ones and zeros may have formed rival street gangs and are refusing to conform to conventional ethics, if you that suits you.)

If you insist on looking for a "practical" answer the likely suspects would be an interruption of power, interruption of the signal path, or an erratic input device.

Assuming that you were using ac power, a little bump to a "weak" cord could cause a momentary drop or surge, or could shake a connector that's not making solid contact. The ground wire, whether separate or just the neutral side of a 2-wire is a critical part of the power circuit. You could also have an internal "loose wire." You might find any weaknesses here by fairly vigorously shaking the wiring while someone plays the keyboard. (That includes shaking the keyboard, if you want to look for internals.)

If you were on battery power, of course check that the batteries are fresh. A weak set may "play" ok, but drop the voltage under the demand of writing a new recording, which could explain not overwriting something previously recorded.

You could also look at the ac power "source." For some sources, the refrigerator in the next room can cause fairly large "voltage surges."

The same investigative "shake" may find a weak connection in the internal or external signal cabling.

If your keyboard has a "volume control," it may be a suspect in the "input devices" category. Most are potentiometer types of devices, and leaving one set at the same level for very long periods may allow "crud" (the technical name for it) to accumulate adjacent to the wiper contact. When this happens, a very slight movement of the "knob" may cause an erratic, and sometimes rather large jump in level. If the wiper contact is "loose," it may sometimes move "on its own" under slight motions of the keyboard. (As in "building up to the crescendo.") A few full range "wipes" of the knob may clear the wiper path and restore normal function, or may at least reveal a "lumpy adjustment" so you can set the control outside the "bad spot."

Unless the problem reoccurs, I'd dismiss it as a few "juvenile delinquent bits cruising in the data stream."

John