On my Debian machine, files are printed one sided, even though I use
Duplex=DuplexNoTumble in my .cups/lpoptions file. I suppose that this
is due to the following issue.

cventin:~> lpoptions -p print-1
ColorModel=Gray copies=1 device-uri=ipps://print-1.lip.ens-lyon.fr:443 Duplex=DuplexNoTumble finishings=3 job-cancel-after=10800 job-hold-until=no-hold job-priority=50 job-sheets=none,none marker-change-time=1631710524 marker-colors=#00FFFF,#FF00FF,#FFFF00,#000000,none marker-high-levels=100,100,100,100,95 marker-levels=25,31,30,72,0 marker-low-levels=5,5,5,5,0 marker-names='Cyan\ TK-8335C,Magenta\ TK-8335M,Yellow\ TK-8335Y,Black\ TK-8335K,Waste\ Toner\ Box' marker-types=toner,toner,toner,toner,waste-toner media=iso_a4_210x297mm number-up=1 output-bin=tray-1 print-color-mode=color printer-commands=none printer-info='print-1 (Kyocera TASKalfa 3253ci)' printer-is-accepting-jobs=true printer-is-shared=true printer-is-temporary=false printer-location='Bocal (left)' printer-make-and-model='3253ci - IPP Everywhere' printer-state=3 printer-state-change-time=1631710539 printer-state-reasons=none printer-type=45180 printer-uri-supported=ipp://localhost/printers/print-1 sides=one-sided

See the "sides=one-sided" at the end, which doesn't come from
my config, where .cups/lpoptions contains:

Default print-1 ColorModel=Gray Duplex=DuplexNoTumble

Ditto with cups 2.3.3op2-7.

Note that I had added this printer with:

# lpadmin -p print-1 -v ipps://print-1.lip.ens-lyon.fr:443 -E -m everywhere \
    -L "Bocal (left)"

On a Debian 10 (buster) machine, with the same .cups/lpoptions:

patate:~> lpoptions
ColorModel=Gray copies=1 device-uri=ipps://print-1.lip.ens-lyon.fr:443 Duplex=DuplexNoTumble finishings=3 job-cancel-after=10800 job-hold-until=no-hold job-priority=50 job-sheets=none,none marker-change-time=1631709869 marker-colors=#00FFFF,#FF00FF,#FFFF00,#000000,none marker-high-levels=100,100,100,100,95 marker-levels=25,31,30,72,0 marker-low-levels=5,5,5,5,0 marker-names='Cyan\ TK-8335C,Magenta\ TK-8335M,Yellow\ TK-8335Y,Black\ TK-8335K,Waste\ Toner\ Box' marker-types=toner,toner,toner,toner,waste-toner number-up=1 printer-commands=none printer-info='print-1 (Kyocera TASKalfa 3253ci)' printer-is-accepting-jobs=true printer-is-shared=true printer-is-temporary=false printer-location='Le Bocal, on the left GN1.347' printer-make-and-model='3253ci - IPP Everywhere' printer-state=3 printer-state-change-time=1631709892 printer-state-reasons=none printer-type=12412 printer-uri-supported=ipp://localhost/printers/print-1

i.e. without "sides=one-sided", which is OK.

So there has been some regression in recent cups versions.

[...]

I had kept the output I got after this printer was installed:

cventin:~> lpoptions -p print-1
ColorModel=Gray copies=1 device-uri=ipps://print-1.lip.ens-lyon.fr:443 Duplex=DuplexNoTumble finishings=3 job-cancel-after=10800 job-hold-until=no-hold job-priority=50 job-sheets=none,none marker-change-time=1602254655 marker-colors=#00FFFF,#FF00FF,#FFFF00,#000000,none marker-high-levels=100,100,100,100,95 marker-levels=96,96,96,95,0 marker-low-levels=5,5,5,5,0 marker-names='Cyan\ TK-8335C,Magenta\ TK-8335M,Yellow\ TK-8335Y,Black\ TK-8335K,Waste\ Toner\ Box' marker-types=toner,toner,toner,toner,waste-toner number-up=1 printer-commands=none printer-info='Kyocera TASKalfa 3253ci' printer-is-accepting-jobs=true printer-is-shared=true printer-is-temporary=false printer-location printer-make-and-model='3253ci - IPP Everywhere' printer-state=3 printer-state-change-time=1602254680 printer-state-reasons=media-empty-report printer-type=45180 printer-uri-supported=ipp://localhost/printers/print-1

That was on 2020-10-09, and there was no "sides=one-sided" at that
time, and this was with cups 2.3.3-3.

So, it is possible that the issue appeared after the switch to the
OpenPrinting CUPS fork in November 2020.

Additional information:

On both Debian 10 and Debian unstable, I get

$ ipptool -tv ipp://localhost/printers/print-1 get-printer-attributes.test | grep sides-
        sides-supported (1setOf keyword) = one-sided,two-sided-long-edge,two-sided-short-edge
        sides-default (keyword) = one-sided

I suspect that for some reason, the old client did not take
into account sides-default, but the fork does. I don't know
whether this is intended, but I couldn't find any information
about such a change.

I can confirm this bug on 2.4.2-3. I am not able to print double sided
even if double sided is selected.