I really wish people would stop calling FPGA's ASICs. Baikal (so far) has only made FPGA mining rigs. There is a huge difference between an FPGA and an ASIC. An FPGA is not that much different from a GPU. Anyone can get one. Anyone can buy a Zynq FPGA board from digikey for $89-$199, or a higher end one for more money, and if you take a little while to learn how to program it, you can hash any algorithm except equihash & ethash. Furthermore, your ROI will be better than a GPU in almost every case, in some cases dramatically better (as Baikal showed with the X10 and Giant-B).
Are you really sure that Baikals X10/B are FPGA? (Which FPGA, by the way?) While it is enterely possible, technical characteristics of these products are not typical for FPGA. Especially the low power consumption.
As for ROI I wouldn't expect good ROI from entry and mid-range FPGA boards because of their weak power supplys. We need propertly designed professional grade DC/DC for core voltage, that is rarely seen in practice.
There is nothing suspicious about the Giant-N. I was already working on an FPGA cryptonight miner before the Giant-N was announced (and obviously I am now focusing on other algorithms). The power of 60W is realistic for one FPGA accessing many external SRAM's. Unlike DRAM, SRAM consumes very little power. The fundamental nature of Cryptonight is that it uses almost no number crunching (by design). A single FPGA just accesses many parallel SRAM's and these memory accesses do not consume a great deal of power. FPGA's consume way less power than other mining devices already. Consider the X10 burns 250-500W and makes the same amount per day as a 2000W GPU rig. Some algorithms burn more, some burn less, and algorithms that have no number crunching (like Cryptonight) burn the least. The reason a Vega 56/64 burns so much power on cryptonight is because it is using high bandwidth external memory, a totally different approach than using many SRAM's in parallel.
FPGA's can be reconfigured very quickly. It is true that certain PCB designs and part selections are better at some algorithms than others. But it doesn't matter if Monero does a hard fork, you can still just use an FPGA to mine the new algorithm, ad infinitum. As I mentioned before, only Ethash is truly resistant to FPGA's. As Baikal has more and more FPGA miners on the market with different types of FPGA's and RAM (Giant-B, X10, Giant-N), a coin which 'forks' would have to know the exact internal configuration of every FPGA mining rig on the market to 'avoid' a new algorithm which could be efficiently mined by them. To give an example, there is a decent chance that Monero's new algorithm could be (accidentally) mineable by the Giant-B or Giant-X10 or Giant-N, and all Baikal has to do is release new bitstreams (firmware for the SD card) that would update those rigs to mine the new algorithm.
As the number of different FPGA rigs on the market continues to increase, it would be very difficult to fork to an algorithm that would be immune to those rigs, unless you pick an Ethash style algorithm. Furthermore if you add in all the cheap FPGA boards available from companies like Digikey, Avnet, Xilinx and Intel, then there is ALREADY a mass produced FPGA board that can do any algorithm efficiently except Ethash.
FYI the Monero ASIC statement is specific to ASIC's. They specifically say they want to avoid ASICs mining their coin (they speak of FPGA's more favorably, and separately from ASICs). Since the Giant-N is an FPGA rig, it doesn't actually fall into the category of something they would fork away from. Furthermore, the Giant-N hash rate is not devastating to GPU's. It has a slightly better ROI than Vega's, but in no way do Vega's become obsolete. Baikal would have to ship out 100,000 Giant-N's to truly disrupt the Cryptonight networks, which is unlikely.
(BTW I bought 2 Giant-N from a local reseller in Vancouver. The units are supposed to arrive on Monday.)
