The following issues arose from testing of the following ICU change to make the 'j' skeleton character mapo to the locale's timeData preferred cycle, as specified by CLDR: https://unicode-org.atlassian.net/browse/ICU-20383
fr_CA - timeData preferred=h, I am pretty sure that is wrong. The time patterns all use HH. => fix timeData.
af_NA. for NA timeData preferredf=h. af_NA has no time formats, for these falls back to af which is for ZA (for which timeData and formats have H). => add time formats with h in af_NA
ckb_IR. for IR, timeData preferredf=H. But default calendar is persian, and ckb_IR only has time formats for gregorian (using H as expected). For persian tiime formats, fall back to ckb, which only has time formats for gregorian (with h) so those get used. => in ckb_IR add time formats for persian, using H.
ha_GH. for GH timeData preferred=h. ha_GH has no time formats, for these falls back to ha which is for NG (timeData preferred=H), which also has no time foirmats and falls back to root formats using H (ok for NG).. => add time formats to ha_GH using h.
lkt[_US] time formats using h are unconfirmed, so in ICU these fall back to root formats using H. -=> confirm lkt time formats (and date formats) same as en[_US].
en_001, ar_001 = ar: timeData for 001 has H (makes sense) but short time formats for en_001 and ar use h. What to do? Changing timeData for 001 to h seems wrong; changing en_001 to use H might be better wbut would have compatibility implications.
en_BI, en_NG: neither has time formats, for both timeData has H, both fall back to en_001 which has formats with h. => add gregorian time formats in en_BI, en_NG, using H.
There is also something strange going on for ar_SA in ICU DateTimePatternGenerator with inheritance for availableFormats time items, but I think it can be addressed by adding availableFormats items for H and h to ar@calendar=islamic formats.