I think the performance of the J20 against PLAAF legacy fighters might have prompted a significant rethink within PLA top brass on the value of 5th gen fighters (and probably also downgraded their valuation of 4th gen by extension).
The huge numbers of F35s the US and allies are procuring would also be a consideration. The PLAAF might expect a J20 to be able to easily deal with an F35, but when F35s outnumber J20s several to one, that changes things since quantity has a quality of its own, as old Stalin used to love to say.
In that respects, I could see the PLA going with a dual track procurement approach.
The first phase would see the PLAAF procure a significant number of carrier capable J21s to operate from land based airfields fo make up numbers quickly, potentially with PLANAF pilots transferring over to the PLAAF to pilot them.
As J20 numbers build up and new PLAN supercarriers come online, those J21s and their pilots would be transferred back to the Navy and transition to carrier squadrons.
This means that in the medium term, the PLAAF builds up its inventory of 5th gens twice as fast as they could with J20s alone, and in the medium to long term, it means the PLAN’s new nuclear powered supercarriers become fully combat capable in the least amount of time possible with pilots who have already had years of operational experience on their 5th gen fighters. It should also not blow up the budget since in the long run, the total number of J20 and J21s produced should not be significantly more than what would have been planned for had the PLAAF stuck to just the J20 and the J21s were only procured for naval carriers.
On the whole, I do not really see the PLAAF having massive appetite for J21s for the same reason China isn’t unduly worried about hostile F35s at present - geography and range.
PLAAF land based J21s would need land bases to operate from, and there is just no one China needs land based J21s to defend against.
Maybe in the long term, as the belt and road really takes off there might be a need for PLA forces to secure it against the inevitable hostile actions, but those would look to be mostly COIN and ground forces work, since no one would be stupid enough to wage a full scale way against a fully connected Belt and Road country as that would be equivalent to starting a ground war with China. In which case legacy fighters and drones would be the workhorses, and any shock and awe intimidation ops can be carried out by J20s.
I still think the J21 is going to be a mostly naval and export fighter. Which maybe why the Chinese initially stay mum about their plans to ultimately transfer their land based units to naval carriers.
But in developing it and having the PLAAF initially operate it should help it to secure some decent export orders. Pakistan would be interested, and China might even clear it for Iran if relations with America and Israel continue to deteriorate. Hell, even the Russians might be interested if there is meaningful ToT, which might be on the cards.
I don't see any rationale for why they would deploy carrier based aircraft on land.
For the PLA as a whole, a J-XY that is produced and in service -- whether it is based on a carrier or based on land -- it is still only one 5th generation fighter.
More importantly, given the development time that the standard carrier based J-XY will take (assuming a maiden flight this year), and going by the pace at which the 003 carrier will be built/launched/commissioned and even accounting for additional 003 pattern carriers that may be built this decade, I do not see any benefit in "delaying" the introduction of carrier based J-XY aircraft to operate them mostly from land before operating them primarily from carriers.
The question IMO is about whether there will be a carrier based variant (that we have all been expecting as the first baseline variant) and whether there will also be a land based variant (which over the last six months we have been getting strong hints of), and what the total number of J-XYs will be built (carrier and land based variants combined) in relation to J-20s.
I agree with you that the PLA has a significant need for a large number of 5th generation fighters, however I think that requirement is likely better served through production of a large run of J-XYs separated between a carrier based variant and a land based variant, in addition to J-20s.
I also believe that the PLA very much does consider the F-35 a threat that deserves to be honored, and for all of the long range strike systems the PLA has to strike the airbases the F-35s operate from in the region, and for all the ways in which they could try to counter tankers that F-35s may depend on for greater combat persistence -- I still believe the PLA would like to have at least parity in terms of deployable 5th generation aircraft from potential adversaries in the region, if not quantitative superiority in the region.
I could envision an ultimate J-20 production run of some nearly 500 aircraft (ending in the mid 2030s), a production run of some 300+ carrier based J-XYs (beginning 2025 and ending after mid 2030s), and a production run of some 600+ land based J-XY (beginning mid/late 2020s and ending 2040).
This of course would be complemented by initiation of production of a 6th generation fighter in the early to mid 2030s (likely by CAC), as well as an unknown number of various MUMT/A2A/loyal wingman, and multirole UCAVs in both the land based and carrier based domains.
In addition, the land based J-XY variant would likely form the basis of an export variant -- i.e.: taking the same airframe but fitting it with export clearance avionics, where the avionics suite will likely prove the biggest "difference" between a domestic PLA variant aircraft and an export variant of the same aircraft.
I could very much see a land based J-XY variant having a larger production run -- potentially double the size -- than the carrier based J-XY "baseline" production run. Both the carrier and land based J-XY variants would enjoy the same avionics suite, weapons suite, engines, and software upgrade spirals, however the difference would primarily be their structural differences in terms of provisions for carrier operations.
Given the numbers involved for carrier and land based variant production for their given airframes, the benefit of having a dedicated land based variant without the complexities of carrier modifications that introduces additional production costs, performance costs and operating costs, would begin to be apparent and worth the development time for a dedicated land based variant.