Jon was using an Elecraft K3 at 100 watts on 20m with a vertical antenna. The recording was made on 27 May at 0713Z.
The delay was 1.272 seconds.
(UPDATE: As Sverre mentions in the comments, it is meaningless to attribute this degree of accuracy to a transceiver - after all, it is not a piece of test equipment. So let's assume a 1.2 second delay).
This delay is too long to be explained by normal long-path propagation of his own signal back to him - over 10x too long.
Jon shared an MP3 file of one of the examples. I opened the file in Audacity so that it could be seen visually and so that cursors could be placed in order to measure the exact delay:
Sverre LA3ZA explains part of the delay here.
But still - almost 1.3 seconds is a very long delay, even when factoring in sidetone/RF transmission timing differences + receive processing time + LP propagation. The sum total should amount to far less than 150mS.....
.
.