Does it really keep the blockchain size in check? Either storage costs shrink at an exponential rate, in which case the block size basically doesn't matter at all, or they don't, in which case the blockchain will eventually grow too large anyway.
The SegWit2X non-debate is never really about keeping the blockchain size (i.e. transactional history) in check as that could be reconstructed relatively easily on demand. However the UTXO set (list of all unspent transaction outputs) has to be kept small because every incoming transaction has to be validated against this ~3GB data set that has been growing by several hundred megabytes per year. Ideally this database should be kept in memory for maximum efficiency but RAM is expensive, so a elaborate cache scheme had to be developed to make sure that nodes remain functional on low power hardware with limited memory and I/O which in turn limits the number of SPV clients each node could serve. Thus on-chain transactions has to be kept to the minimum to slow the growth of UTXOs and SegWit provides some incentive for people to consolidate dust outputs despite fee pressure. So far this strategy has not been working.
Ethereum does not have this limitation because balances are kept in individual accounts, thus the requirement for global consistency is a lot lower.
It is a lot for embedded and mobile devices, even the ability to address 4GB memory adds to the cost of processor. Right now full history nodes on phones are a pipe dream, and people who run full nodes on raspberry pis are actually degrading the performance of the network by adding a tonne of slow nodes with 1GB memory or less.
In the long run the UTXO set will always grow because of lost keys and dust amounts worth less than the fees to move them, yet it is politically unacceptable nor practical to drop any of these from the consensus, so the problem will only get worse.
It is, of course, not a problem if you think SPV wallets are the way forward, but this view has been stigmatised as "centralisation" so it also leads to no real solutions.
