True, hosting costs don't increase with Dash price but they do and have increased as Dash masternodes need to provide more services. The release of Dash Platform will also increase hosting costs and as Dash Platform is adopted, hosting costs will likely have to continue to increase.

Now, let's just illustrate reality. Instead of 2015, let's use near end of 2016, which I contend matches up to where we find ourselves right now within Bitcoin's 4 year market cycle.

2016: masternode count = 4100, paid roughly every 7.38 days or just over 4 times a month, block reward = 1.8, total reward = 7.3 DASH @ $9.50 each
2020: masternode count = 5000, paid roughly every 9 days or 3.33 times a month, block reward = 1.44, total reward = 4.8 DASH @ $65 each

2016: hosting cost per month = slightly more than 0.5 DASH at roughly $5
2020: hosting cost per month = slightly less than 0.5 DASH at roughly $30

2016: net DASH per month = 6.8 DASH and 27880 DASH across all masternodes
2020: net DASH per month = 4.3 DASH and 21500 DASH across all masternodes

2016: total DASH = 6.95 million, 4.1 million in masternodes which is 59% of supply in masternodes
2020: total DASH = 9.80 million, 5.0 million in masternodes which is 51% of supply in masternodes

Ok, so let's check some of the things you said.

1) You said more DASH is at $0 base cost now than before. And even if I accept your argument that DASH masternode rewards are $0 cost base this is not true. Each masternode receives about 36% less DASH than it did roughly 4 years ago and the total that all masternodes receive is just under 23% less.
2) You said it's harder to create masternodes today because the supply is more fragmented. Yet we have 8% more supply that's not contained in masternodes that can be bought than roughly 4 years ago and can be bought at a cheaper rate with top coins BTC and ETH. From 2016 it seems masternode count added another 500 before DASH topped out in price around mid December 2017. From today starting with 8% less DASH tied up in masternodes, just to match the ratio 4 years ago, we'd need to add another 780 masternodes. Also, consider the supply increase that occurs as blocks are mined throughout 2021 and the real possibility that StakeHound is even moderately adopted. All that indicates to me that we have more than enough runway to add another 500-1000 masternodes.

As illustrated conditions leading up to the 2017 bull run seem to be in place for the 2021 bull run. Good luck and just remember this time to plan properly and take profit when you have a chance.