So no substitute worth teensy 3.6 before 2020?
Well no, probably not.
In many ways this could be seen as a replacement or upgrade. This new mid-range ~$24 product has nearly all the features of today's high-end $29 product, plus ~4X faster performance, much more memory, and some pretty awesome new peripherals.
But in some other sense none of the new products directly replace or substitute for the old... unless NXP manages to give us extremely unlikely silicon next year. Implementing 5V tolerance and higher quality analog is very unlikely due to the physical realities of faster process silicon nodes. NXP's Kinetis chips on Teensy 3.x use a 90 nm process. These new iMXRT chips are somewhere between 28 to 40 nm (haven't gotten a clear answer - this may be something they're just not willing to say anymore). The silicon is physically different and just can't do some of the analog stuff the old parts could - so not a direct substitute.
2020 probably sounds like a long way in the future, since we're only at the end of 2018 right now. But again, it's important to keep things in perspective. Right now, they've not even publicly announced any parts beyond the RT1064. Usually info first appears on their website and only limited specs in the 2-page "fact sheet" PDF. Typically months later they actually publish the reference manual and some time later early samples become available. Usually it's still more months until the parts actually become buyable in volume. Then there's the lead time to actually get them, which is currently ~14 weeks. But that's all for these current parts which are basically the same as the RT1052.
Consider how long the RT1052 took. I was told of the chip (under NDA) in 2016. They published the first datasheet & manual (under NDA) in June 2017, and put the first "fact sheet" on their website around that time. Later in the the summer of 2017, the first eval boards shipped (also under NDA) and became publicly available near the end of 2017. But that was the early version of the chip which had severe errata. It was only months ago that the version with full 3.3V power supply compatibility and atomic bit set/clear GPIO became available.
Hopefully NXP will have a smoother release of whatever next-gen chip comes, well, next. That is, assuming they are going to make add more chips in their iMXRT product line. Anything I might know of their future plans would be under NDA until they actually publish info on their website. But if you carefully compare their many products, it's pretty clear to see the RT1052 is pretty much ARM's Cortex M7 with a mix of stuff from iMX 6 and the newer Kinetis parts, plus some stuff maybe custom (FlexRAM, DC-DC switcher, etc). It doesn't take much imagination to look at their current iMX 7 & 8 product lines to see the pretty incredible IP they've already developed and shipped on high end application processors and hope we might get some of that amazing stuff in future iMXRT. There's also some change they'll surprise everyone with totally new stuff, like when the first Kinetis chips with FlexIO came out a few years back (right around the time Teensy 3.6 was releasing - about 1 year after the K66 chip came out).
The other reality to consider is PJRC is a small shop with limited resources. I do not believe we've ever managed to release 2 new Teensy models in the same year. So if 4.0 (likely RT1062) and LC2 (maybe RT1015) both make it to market in 2019, that'll be a first for us. Also notice NXP's website says RT1015's status is still "pre-production" and you can't yet download a reference manual or datasheet. It's very unlikely that chip will actually be available in volume until mid-2019 or later.
Consider the time it's taken Arduino to release new products using a substantially different chip. I believe Leonardo (USB AVR), Due (SAMX) and Zero (SAMD) each took well over a year to come to market (and at least with USB AVR they had Teensy 2.0 available for 3 years for inspiraton....). Even though they're much bigger with more developers, this sort of work to support a new architecture simply takes time. But something we're doing differently than Arduino, which can make this seem much longer, is giving early access and input on the design via the forum. We try to balance this early access with a few politely requested restrictions (mainly not showing photos during beta testing). Often I've wondered if this has been unwise. Certainly one of the downsides to such early access is it really shows the full span of the very long development cycle. Please do try to keep all this in perspective...