Hello Badunit,
I'm not sure I understand your example correctly, but it appears D6 and D7 show strings with 3 components of date which can be interpreted as dates without ambiguity, in which case SUMIFS() in D8 returns the result we expect.
The question is the format of dates in D6 and D7 and not the format of dates in A. SUMIFS() accepts values in A as independent parameter and so it can honour the original type of values in A, while it accepts values in D6 and D7 as part of string which is interperted as condition and thus it cannot know the original type of values in D6 and D7 but only re-interpret their current string values which are affected by their current format.
This issue is not limited to date/time values. Number format of cells used in condition part of those functions also affects the result.
E.g.
Case 1 as we may expect:
A2 1.001
B2 1
C2 =A2 # 1.001 (formated as 3 decimal places)
D2 =SUMIFS(B2:B2,A2:A2,"<="&C2) # 1
Case 2 as we may not expect:
A4 1.001
B4 1
C4 =A4 # 1.00 (formated as 2 decimal place)
D4 =SUMIFS(B4:B4,A4:A4,"<="&C4) # 0
Case 3 as a workaround for Case 2:
A6 1.001
B6 1
C6 =A6 # 1.00 (formated as 2 decimal places)
D6 =SUMIFS(B6:B6,A6:A6,"<="&(C6+0)) # 1
Formula in D6 in Case 3 yields the desired result 1 regardless of the format of C6. This is analogous to the previous workaround for date using, e.g., ">="&(A1-NOW()+NOW()) in lieu of ">="&A1 where A1 contains date with arbitrary format.
I'm yet to convince myself to consider it bug because it is the expected behaviour of & operator which honours the format:
A8 1.001
C8 =A8 # 1.00 (formated as 2 decimal places)
D8 =""&C8 # 1.00
However, I do agree that it is defect of condition parameter of those functions such as SUMIF(), SUMIFS(), COUNTIF() and COUNTIFS().
All the best,
H