i've got an enquiry out with Linear, however it looks like there's a serious design flaw with the LTC4155 - not one that's safety-related but one that makes it functionally completely useless.
the clue is in looking at page 15 of the datasheet - a diagram which shows the internal operation. in the bottom left corner is a power-selection arrangement, where there are two AND gates. note that they're not symmetrically arranged. the difference between the two is that WALLSNS output is connected to the *input* (inverted) of the USBGT enabler AND gate.
what that means is that it is ABSOLUTELY CAST-IRON GUARANTEED - i.e. hard-wired in an inviolate way - that if the 5VDC input is plugged into a wall charger, USB OTG Host Power is ABSOLUTELY 100% WITHOUT FAIL ABSOLUTE guaranteed to be terminated without warning.
which:
(a) violates the USB-OTG Power Specification (b) is in direct contravention of the statement on page 1 of the LTC4155 datasheet which *clearly* states that the LTC4155 is compliant with USB-OTG Host Power Mode (c) is in clear violation of various laws like "Trade Descriptions Act" of the UK due to the qualifier statement in paragraph 2 which reads "....without needing additional components".
all of which is a frickin nuisance, as i now have to go find alternative dual-supply OTG-compliant [boost-mode] battery charger ICs. i've found two already - bq24192i ($7 in low volume and it would require a 2A 5V step-up converter IC as well....], or the MAX77818 which is a 0.4mm pitch BGA - perfect because it even has battery status monitoring built-in, but the 0.4mm pitch BGA _wow_ that's gonna be tough for small china pcb assembly factories to deal with when they're used to QFNs and QFPs.
argh!
i'll let people know how this goes: i'm expecting to hear from Linear, they *might* confirm that the I2C "REQUEST_OTG" mode does in fact honour the USB OTG Host Power mode properly. if they do, then all is well - otherwise i will need to do a redesign of PCB3.
l.