looking for best end-to-end workflow from video editor to dash abr for VOD

I'm looking to identify the best workflow from my video editor exports (from premier pro) to a set of segmented dash videos ready for typical web-based VOD streaming up to 1080 HD -- so the output needs to be the typical abr bitrate/rendition ladder that works well with the typical wide range of devices used for web viewing. My raw footage used on my editor timeline was recorded in 4K log format from a high quality drone camera, so the bitrate of these clips is very high (200-300mbps) with a correspondingly large file size that is 10x what a reasonable 1080 HD h.264 file would be with, say, 8mbps bitrate targeted for normal web consumption. So, obviously I need to export reasonably sized file(s) from Premiere to use as a source for some encoder/packager (ie, mediaconvert or mediapackage).

I see these options and related questions:

1. MEDIACONVERT: Export a single media file from premiere to use as an input for mediaconvert and the use either automated or manually defined abr dash renditions from mediaconvert. The question here is, since Premiere can export just about any format at any bitrate and any type (vbr,cbr, qvbr), what would be the best type of export from premiere to use as the input for mediaconvert? My first attempts use h.264 1920x1080 12mbps vbr mp4 for the export from premiere, similar to what youtube recommends to feed its abr transcoding for 1080 HD (max resolution) video ladders with around 6mbps max bitrate -- results appear to be OK, but... is h.264 the best input format for mediaconvert given the output will be dash? Would mediaconvert work better with CBR vs VBR input files? What amount of additional input file bitrate is enough (compared to the highest bitrate output rendition) to ensure no significant loss of quality when transcoding to dash?

2. MEDIAPACKAGE If I know the abr encoding ladder I want in terms of renditions and bitrates, would it be better quality-wise to export all the various rendition/bitrate combinations directly from premiere as, say, h.264 MP4, and then simply let a transmuxer, like mediapackage, repackage the renditions into a segmented dash package without any need for re-encoding --- ie, avoid the losses of re-encoding the premier exports like mediaconvert would do? One concern here might be that creating dash segments (eg, 4 seconds) from a single source file that is not being reencoded, just repackaged, might be a problem since it seems mediaconvert (option 1 above) creates the segment durations intelligently in terms of aligning GOPs with segment boundaries, etc, whereas I assume mediapackage would not be able to do this when just re-packaging.

Other options?

I see very little guidance out there on these aspects for this kind of end-to-end workflow -- most discussions/docs assume that the "input" file just exists, with no guidance on what that preferred characteristics of the input file would be, especially for my type of scenario which I assume is fairly common. Any insights/recommendations on the pros/cons of options 1 and 2 above, and the sub-questions mentioned, would be greatly appreciated.