I'm thinking it may be you as the problem. You crossing any unusual digits during composition? Have you checked your midichlorian levels?
Well, I did go through Ki development training in the early 70's (it's an Aikido thing) -- when Star Wars first came out in 1977 and Obi Wan was describing the Force to Luke, I immediately recognized it as Ki (translated as "life energy" or "mind"; transliterated from Chinese martial arts mysticism as "Qi", though it's written almost the same in both Chinese and Japanese with the rice-like radical in traditional Chinese being simplified into a kind of X)). But unlike in Scientology we didn't check any pseudo-scientific "midichlorian levels", but rather we would use Ki directly in our techniques which depended on us using Ki -- in Aikido, if you try to use physical strength to muscle your way through a move, then it will fail.
I also never crossed the streams (nor even realized why the very thought of it would freak out Ray Stantz so very much until two decades later).
Now, remember that I'm a retired software engineer with initial training as a computer electronics technician (which included chasing sparks through a CPU's logic diagrams). Combine my hardware training with working in C, which allows you to work very close to the metal, and I was very comfortable working with low-level operations, including bit-fiddling.
On my own I had learned sockets programming, which is how network programming is commonly done, about 20 years. While sockets supports upwards of 25 different networking protocol families, TCP/IP is the one used on the Internet and hence the predominant one. So in the process I also learned what I could of TCP/IP and its various protocols.
General networking theory is based on the OSI Protocol Stack which consists of 7 layers (hence the discussion of "layer"). TCP/IP commonly combines some of the OSI layers and is described with four layers which I discuss on my TCP/IP page.
As for HTTP message fields, ever since the mid-80's, I've had a personal programming project that I keep restarting every time technology changes. In 1981, Navy Reserve officer Larry Bond published his modern naval warfare miniatures game, Harpoon (he had designed it to help train fellow officers) and around 1985 a friend and I sat down to run a very simple introductory scenario involving a Kidd-class destroyer and three Osa-II missile boats. The engagement was for 10 to 20 minutes, but after three hours of graphing, measuring, and calculating by hand we weren't even half-way through the scenario. That's when I decided that I needed to develop a "Harpoon Helper" program that would do all that messy book-keeping work for us. Yes, Bond's game system has been converted to a series of computer games, but there your forces are pitted against the computer's and (in the first iteration at least) you have to micro-manage everything.
In the most recent development effort (circa 2013), I expanded my ideas to a networked system. That was inspired by the Distributed Interactive Simulation (DIS) standard in which each unit (including each piece of ordnance in flight) is a separate entity that communicates with all the others (and with the simulation monitor) via UDP datagrams. However, DIS uses multicast and the source code was mainly interface shells, but it got me thinking.
I chose to model the application protocol after the HTTP messages that the Web uses. I studied HTTP and started some prototype coding to process my version of HTTP messages. Again, HTTP request and response messages are how web pages operate as they GET resources and then receive them.
BTW, Larry Bond had worked with Tom Clancy on Clancy's break-out novel, "The Hunt for Red October", for which they used Harpoon to game out the "battle scenes". A major source of military hardware information that they used came from the well-known and authoritative Jane's books started by Fred T. Jane as reference books to support his naval war game (1906). Wargaming is also very much a part of the curriculum at the Naval War College -- Invicta's videos, eg at How Does the US Navy Run Wargames? - A Guided Tour of the US Naval War College Wargaming Facilities
But ever since I retired three years ago, I've done almost no programming at all. I will undoubtedly never complete that project, but it's been a very interesting study.
Fun story. At my last job (my final 28 years of employment), because of my skills in bit-fiddling I would usually be given the task of interfacing our units with new GPS receivers, with the communications protocol of their serial comm ports.
Part of that was interpreting binary data fields. Most receivers would transmit their binary data using IEEE standards, so the biggest problem would be translating the data stream's big-endian data representation to Intel's little-endian data storage. Fairly trivial.
However, we started using a major GPS standard which used an entirely different format for floating-point data. It followed the same basic principles of IEEE-754, but the exponent and mantissa fields were different sizes (which also affected how to interpret the exponent field), along with being big-endian. Plus we had to be able to convert both ways, from RCVR (receiver) to IEEE as well as from IEEE to RCVR. The problem was definitely non-trivial are required concentration and a lot of attention to detail. One flipped or misplaced bit (out of upwards to 64 bits) would be disastrous.
At the same time, I assisted a West Coast Swing dance teacher with her classes, mainly minding the door, handling the sign-ins and the money, and demonstrating moves. One of her popular classes was ladies' styling. It didn't take me long to figure out that it wasn't polite to be the only guy present with nothing to do but sit and watch, so I developed the habit of always bringing something to read or to work on.
Such as this non-trivial floating-point format conversion problem from work. During one 1.5-hour class, I worked out the entire approach, then refined the approach, then sketched out most of the code. The next day at work all I had to do was type in the code, put on the finishing touches, and test it. Worked fine, so I submitted it for code review. I had mentioned where and under what conditions I had worked on it, so everybody, assuming that I had to have been distracted, did their best to find something, anything, that I had done wrong. They couldn't.
ABE: Oh, and yes, there must be something about my site, which is where I place my images, that the forum software doesn't like. I even placed an HTML IMG tag for that Vienna French Toast in the page I'm working on now and it displays just fine, but when I copy-and-pasted that tag to this reply as a test it was "Image Not Found" all over again.
Re: What French Toast (Pain Perdu) is Called Elsewhere
I haven't been able to fix this but will continue to look at it. The image displays fine in Firefox, displays displaced to the right in Safari (will be fixed), but won't display in Chrome or Edge. Both Chrome and Edge display the image fine in a separate tab but get an error when they try to load the image for display in an EvC message.
Re: What French Toast (Pain Perdu) is Called Elsewhere
Just figured it out. It's a security issue. Chrome will not display http images on an https page. The problem should have begun for you as soon as I moved the website into the cloud, because I changed it from http to https at the same time. That would have been January of 2019.
I know how to work around this and should have it fixed soon.
If it's possible to flip your site into https mode, that would also fix the problem.