I don't see heat dissipation as an issue in the bundle itself if the bundle is 24" or shorter. . The "maintaining spacing" concept could be a problem with 15 HRs in constant contact for their entire length thru the attic. . But you have to define "maintaining spacing". . If you choose the wording of 392.11(A)(3) then if you have more than one layer or have less than one cable width of space between any 2 cables, you are not "maintaining spacing".
You would then have to derate according to T310.15(B)(2)(a) and limit your 14 & 12 gauge Romexs to bundles of *4* Romexs if you didn't want to reduce your breaker size.
I'm still not seeing the limit of 3.
Here's the math:
*4* 2wire Romexs are 8 current carrying conductors.
*4* 3wire Romexs are still just 8 current carrying conductors because you don't count the noodles [310.15(B)(4)].
8 CCCs have a 70% adjustment factor [T310.15(B)(2)(a)]. . The starting point for 12gauge is 30a [T310.16 & 334.80]. . 30a x 70% is 21a. . Since you're still above 15a [240.6(A)], you can still use the 20a breaker [240.4(B)] the same as without adjustment [240.4(D)].
*5* 2wire Romexs are 10 current carrying conductors.
10 CCCs have a 50% adjustment factor [T310.15(B)(2)(a)]. . Again starting point for 12gauge is 30a [T310.16 & 334.80]. . 30a x 50% is 15a. . Since you're at 15a, you have to use that breaker instead of the usual 20a.
If there’s a fire stopping bundling issue, at first glance you might think you’re limited to *2* Romexs based on how the second paragraph starts. . But 334.80 ends with the words “the allowable ampacity of each conductor shall be adjusted in accordance with Table 310.15(B)(2)(a)” which brings you right back to the math.
If there’s a bundling issue, you’re limited to *4*.
Where’s the limit of *3* come from ?
David