The only "issue" with retaining compounds ( like 638/648 ) and BB86 is that none of them work with delrin. Alu/steel/Ti to CF -> yes, delrin... no.
If BB shell is out of specs ( delrin is not compressed enough and relatively evenly ) then frame is ... useless. Only solution are brackets from companies like Rotor, Kogel or C-Bear ( alu cups ). Maybe some day some business will do "oversized" delrin cups for worn shells... (?). Compound is able to "cover" up to 0.25mm( 638) or 0.15mm (648) max. For well worn shell it can be not enough.
Bonding Delrin (acetal) is a lot like bonding polyolefins. It needs a specific primer and adhesive but with those it can be bonded quite well. Page 17 of this design guide (https://www.ellsworth.com/globalassets/ ... onding.pdf
) recommends Loctite 770 primer and 401 Prism adhesive. I have used this combination to bond polypropylene-based thermoplastic elastomers and the result is a bond that is far stronger than the rubber. Without the primer, virtually nothing will even begin to bond to the TPE. Per the chart provided in that pdf, acetal seems to behave very similarly.
The only caveat with using 770 and 401 is that you will need act quick. The bond happens in a few seconds but it does take some time to reach full strength. Be prepared with a wood block and hammer to tap the bearings the rest of the way in should they freeze up before fully seated.
As far as filling large gaps, Loctite has a product for that, too: http://www.loctite.sg/sea/content_data/ ... mpound.pdf
I've used that stuff to repair a crankshaft end on my Honda after the crank pulley driving the time belt came loose and wore away a good chunk of the hardened steel. Been going strong for ~8 years since that repair.