After some soul searching, jumping over 5.3 to 5.4 sounds good on paper. However, many products don't work with 5.4 yet, including SquirrelMail. IONCube won't have a release out until next week. The 5.3 migration is a worthy one because you pick up features you can use. 5.4 drops what was deprecated in 5.3, and if that were all, almost all of 5.3 compatible code would run with it, but that is not the case. 5.4 corrects things in the language that were never right from the start. Example: You pass an A-z character or white space to a function, and it evaluates it as 0. You can see where developers would write code that would rely on that behavior. Or lets say you pass a 1.4x, and it evaluates to 1. There are probably a thousand ways hackers can exploit PHP coded sites with a little knowledge of the source code. Suddenly with 5.4, it sees these things as errors. Moreover, it's likely that many problems would be intermittent because whether they show up or not would depend on the contents of a variable at certain point in time. If the code is designed to rely on the new behavior of 5.4, it may not work in <= 5.3 because older versions wouldn't set an error you depend on for branching. Therefore, 5.4 also has a lot of potential to break things. Even after SquirrelMail and IONCube have their releases out, you would have the expense of hiring a body guard if you rolled out 5.4 in a shared hosting environment if the useful plugin scripts that come with products like Installatron or Fantastico haven't been converted yet.
The reason things are moving awfully fast today, is they moved awfully slow in the past. What is amazing to me is why PHP hasn't had a mass exodus. 5.3 was an emergency release of 6.0 without the UNICODE. They recently rebooted the 6.0 project because of a faulty UNICODE strategy. We all agree the changes are necessary, we all agree the changes cause major carnage, and there is really no nice way to do this. PHP was never envisioned to do the things it's being asked to do today.