Without exact project details nobody can tell you anything. The pertinent assumption would be that you are using some compressed format and in order for timewarp to do its magic, large parts of it or even the entire clip need to be fully decoded over and over again instead of just the current frame(s). rendering to a more conventional intermediate format would eliminate that limitation/ bug.
Warp Stabilization, Temporal effects like Time Remapping or Time Warp, and especially temporal effects involving GOP (MPEG) compression should be rendered before you do any other effects. In the long run this is way more efficient. If you have GOP original footage the it is also a very good idea to transcode that footage to a suitable frame based production codec. In my workflow I always transcode any footage that is going to require a lot of color processing or temporal effects to a 10 bit or better production format. Really heavy effects work and compositing is nearly always done using 32 bit image sequences that I create from the original footage. You can easily slow rendering down by a factor of 10 or more by sending the wrong kind of footage to the render cue or by applying temporal effects in the wrong way to the wrong type of footage. Every time I have a render that throws me an estimate I didn't expect like 17 hours I take a look at the project and try and figure out what I did wrong.
BTW, MP rendering with Temporal effects is usually a very bad idea. That alone, combined with a highly compressed original footage may have been 90% of your problem.
More details. Footage:
Format : AVI Format/Info : Audio Video Interleave File size : 591 MiB Duration : 12s 733ms Overall bit rate : 390 Mbps
ID : 0 Format : Fraps Codec ID : FPS1 Duration : 12s 733ms Bit rate : 388 Mbps Width : 1 920 pixels Height : 1 080 pixels Display aspect ratio : 16:9 Frame rate : 60.000 fps Color space : YUV Bit depth : 8 bits Bits/(Pixel*Frame) : 3.118 Stream size : 589 MiB (100%)
Bottom to top:
Layer 1: A clip that was trimmed to 53 frames long.
Layer 2: Adjustment layer: timewarp used to make 4 copies, alternating forward and reverse, .i.e. forward then reverse then forward then reverse. Using source frame keyframes 0/0 52/52 105/0 158/52 211/0
Layer 3: Adjustment layer: timewarp with a complicated bezier adjustment easing in in some places and out in others. The composition is lengthened from 212 frames to 620 frames.
Layer 4: Adjustment layer: timewarp with 3 source frame keyframes 0/0 619/619 668/619
I used precomposition incrementally:
With layers 3 and 4 hidden select and precompose layer 2 and prerender
With layer 4 hidden precompose layer 3 and prerender
Precompose layer 4 and prerender
That took roughly 30 minutes for each layer. That was slow enough. I didn't initially try composing all 3 layers together because I didn't want to wait 5 hours that it took to render the composition without prerendering.
Actually, I started with the footage in premiere, in it's own sequence and used replace with after effects composition. But, the difference is dramatic enough without involving premiere
Ack.. I hope no one reads this before I fix it: Using import and import, I did this:
Footage clip of 53 frames + adjustment layer using timewarp and resulting in 212 frames export as uncompressed AVI. Rendering took 30 seconds.
Importing the AVI and creating a new composition. Adding the complicated bezier timewarp export uncompressed AVI. Rendering took 2.5 minutes
Import the AVI and creating a new compositon. Adding the 3 keyframe timewarp export uncompressed AVI. Rendering took 2 minutes
Importing the final rendered result into premiere took some few seconds.
I can't reproduce the problem.
One thing about this though. If there is a reason to use rendered video instead of precomposing, it's slightly easier to prerender and use the resulting video, discarding the original composition, than it is to render and then import the renderered video.