This is probably a problem which is much more amenable to testing and good sound engineering experience and judgement, than to a specific find of a code paragraph covering it. I doubt that any code would condone this. Although, there will be plenty of evidence, text and research on why this is a difficult problem and thus generally not allowed by the codes. However, most codes do allow you to do things, or use methods and materials, which are not specifically covered by the code, with sufficient engineering proof of the scheme.
3/8" screws into the edge of a ½" web pl. doesn’t allow for much edge distance, and your predrilling, countersinking and tapping operations had better be almost perfect. Even if the drill and tap is perfectly centered on the web, you might get some deformation (bulging) on the side faces of the web. Twice as many 3/16" or 1/4" bolts might be better. I would make sure that the screw or probably machine bolt was the weaker element, the failure mechanism, because of the almost nonexistent edge distance on two sides. Then design the screw spacing, length, etc. accordingly, with an appropriate FoS. You should be able to develop pull out and shear flow type strengths assuming that the loads aren’t too outlandish. You may actually want to, have to, machine the edge of the web pl. for good fit-up and bearing on the flg. pl. before drilling, etc.
What are the loads and how are they applied? Pull-out will be a function of machine bolt length and some torque limit to assure you have good bolt head bearing on the flange, to resist the applied loads. And, prying on the bolts could be an issue. The shear or shear flow problem involves a machining and assembly method which assures, with some confidence, that you pretty much bring all the bolts into play at the same time. The tap for the thread in the web pl., threads not in the faying plane, a counter bore which exactly fits the bolt shoulder and is concentric with the tapped hole, and finally some bolt torque for clamping. This leads to a multi-axial state of stress in the bolts and in the steel immediately around them. The tolerances in typical bolted joints generally causes a few on the bolts to come into play, in bearing on steel pl. holes, and actually causing yielding in bearing, before more of the bolts are brought into play. Thus, 10 bolts in a joint doesn’t mean each bolt is loaded to (the load on the joint/10). Some of these problems tend to go away when using self tapping screw on light gage materials.
The work and expense of what you are trying to do will be a real budget buster if you are going to do it right and with some confidence, and will probably have to be done in a machine shop setting. It is probably not advisable in a field connection situation, where tolerances and control will be lost. I’d work with the client to devise another connection solution, unless his reasons truly justified the costs. You don’t give us a hint as to why your client is making things so difficult, and I wouldn’t stick my neck out for him without some real justification. Whatever your 3/8" bolts at 2" o/c will do, two 1/8 or 3/16" fillet welds would do with far less cost and pain and suffering.
