Here's the skinny on what you are proposing, Aaron...

You can absolutely set up the base with an autonomous position, perform RTK of all points within range of the base radio and later process the base to CORS. The DPOS process in J-Field will translate all vectors from the autonomous base position to the CORS derived position of the base. Depending on your local CORS, you can actually get a solution at the top of the hour after the end of the session. This means if you stop the base at 10:10am, or 10:55 am, you can process at 11:00am, and get a CORS derived base position. This is provided that your local CORS upload hourly. Most do except for the PBO stations which I believe still only upload every 24 hours at 0 UTC.

Now for the things to watch out for...

If your base file gets botched, you're session is lost with no way to recover the anchor point for the session. This is extremely, extremely rare, but a possibility. We have many tools now to recover DPOS sessions, so it is even rarer than it once was. But there is still a chance. A more likely issue is the duration of the observation. More time = More accuracy when it comes to CORS processing. I just completed a survey in which I occupied the same base position on seven different days. I sent each file to DPOS for processing the base>rover and the base>CORS. I held the first day processing for the remaining six sessions. Here is how they differed (all distances are US Survey Feet):

Day 1 (414 minutes)

0 N, 0 E, 0 U (Held)

Day 2 (405 minutes)

+0.003 N, +0.005 E, +0.026 U

Day 3 (332 minutes)

-0.017 N, -0.013 E, -0.020 U

Day 4 (240 minutes)

+0.058 N, +0.083 E, +0.069 U

Day 5 (356 minutes)

-0.013 N, +0.012 E, +0.020 U

Day 6 (267 minutes)

-0.033 N, +0.026 E, +0.020 U

Day 7 (376 minutes)

+0.016 N, +0.009 E, -0.007 U

Day 4 is the shortest observation which serves to prove my point. I would generally expect better results from a 4 hour session, but it is telling that the shortest observation also proved to have the poorest accuracy. Day 6 was only slightly longer (by 27 minutes) and is the second worst accuracy. Better than Day 4, but not as good as the others with over five hours of observations. Of course, your CORS geometry will be different from my CORS geometry, so you may experience better or worse results with similar time spans. If you are looking to have a quick set-up at each base point and tie a few RTK shots, then advance to the next part of the project, I think you may be opening yourself up for diminished accuracy. It depends on your project requirements as to whether the loss of accuracy resulting from short base observations will exceed your error budget. It will also depend on your CORS as to how much accuracy will be lost due to shorter observations.